56
Outsourcing of Grid Computing Master of Science, Systems Science Project Report submitted to Department of Computer Science, Louisiana State University Santiago Pe˜ na * May 23, 2007 * Department of Computer Science, Louisiana State University, Baton Rouge, LA 70803.

Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Outsourcing of Grid Computing

Master of Science, Systems Science

Project Report

submitted to

Department of Computer Science,

Louisiana State University

Santiago Pena ∗

May 23, 2007

∗Department of Computer Science, Louisiana State University, Baton Rouge, LA 70803.

Page 2: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Abstract

This work presents an overview of the outsourcing phenomenon applied to Grid

Computing. Grid Computing is an exciting technology that has been making significant

advances in academia since the development of the Globus Toolkit. In recent years,

the technology has become more intriguing to the business world but there appears to

be a void in the literature regarding the outsourcing of Grid Computing. The dearth

of coverage on the topic is understandable because Grids are a relatively new and

complex technology that researchers in the business field are not entirely familiar with.

Furthermore, Grid Computing has existed primarily in the academic and research

world since its inception, and the technology has not historically been viewed as a

resource/service to be outsourced. The goal of this research is to determine the factors

(e.g., cost, scalability, compatibility, etc), that have the largest influence on the business

decision-making process to outsource Grid Computing.

1

Page 3: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Acknowledgements

I would like to express my sincere gratitude to the following individuals for their contribution

to this work: Dr. Gabrielle Allen, Dr. Rudy Hirschheim, Dr. Christopher White, Charles

McMahon, Mark Hicks, and Prabha Krishnamachary. Additionally, I would like to acknowl-

edge the support from the UCoMS and SCOOP projects and the Center for Computation

&Technology at Louisiana State University.

2

Page 4: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Contents

1 Introduction 5

1.1 Defining Grid Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.1.1 Business Grid in Academia . . . . . . . . . . . . . . . . . . . . . . . . 8

1.1.2 Academic Grid in Business . . . . . . . . . . . . . . . . . . . . . . . . 8

1.2 Outsourcing Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2.1 Defining IT Outsourcing . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2.2 Outsourcing Stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Market Analysis 12

2.1 Grid Solutions in the Market . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.2 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Outsourcing Grids 24

3.1 Why Outsource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.1.1 Cost Savings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.1.2 Technical Reasons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.1.3 Core Competencies Reasons . . . . . . . . . . . . . . . . . . . . . . . 36

3.1.4 Environmental Factors . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.1.5 Innovation Diffusion Theory (IDT) . . . . . . . . . . . . . . . . . . . 37

3.2 What Needs to be Outsourced . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.2.1 Total Outsourcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.2.2 Total Insourcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.2.3 Selective Outsourcing . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.2.4 Joint Ventures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.3 How to Outsource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.3.1 Vendor Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.3.2 Contract Negotiation . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.3.3 Contract Management . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3

Page 5: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

4 Grid on the Bayou 47

4.1 Economic Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.1.1 A Common View of Economic Development . . . . . . . . . . . . . . 48

4.1.2 A Comprehensive View of Economic Development . . . . . . . . . . . 48

4.1.3 A New Structure to Foster Economic Development . . . . . . . . . . 49

5 Conclusion 51

List of Figures

1 Stage model of IS outsourcing - Source: [12] . . . . . . . . . . . . . . . . . . 11

2 Detail of Costs for SuperMike: the power consumption was assumed to be

100kWh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3 Percentages of the cost of ownership per category . . . . . . . . . . . . . . . 28

4 Detail of Cost for SuperMike: the power consumption was assumed to be 80kWh 28

5 Percentages of the cost of ownership per category . . . . . . . . . . . . . . . 29

6 Detail of Costs for IBM P5: the power consumption was assumed to be 35kWh 29

7 Percentages of the cost of ownership per category . . . . . . . . . . . . . . . 30

8 Detail of Costs for SuperMike: the power consumption was assumed to be

30kWh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

9 Percentages of the cost of ownership per category . . . . . . . . . . . . . . . 31

10 Aggregate Percentage of Savings per Category, for cases 1, 3, 2, 6, 8, and 10 32

11 Rogers’ Innovation Adoption Curve . . . . . . . . . . . . . . . . . . . . . . . 38

12 Clauses included in detailed contracts . . . . . . . . . . . . . . . . . . . . . . 45

13 LONI Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4

Page 6: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

1 Introduction

Grid Computing is an exciting technology that has been making significant advances in

academia since the development of the Globus Toolkit. In recent years, the technology has

become more intriguing to the business world [21].

Grids have the potential to be utilized by businesses to dynamically share large amounts of

data amongst themselves and their business units. Examples of such organizations include

hospitals, petroleum companies, research facilities, banks, and even governmental offices.

The ability to share resources (computation and data) could improve the efficiency of such

institutions, translating into an increase in economic benefits. According to the Guide to the

Networking and Information Technology Research and Development (NITRD) committee:

“. . . grid computing is becoming a key strategic element in the alignment between business

and IT.” Consequently, it is not surprising that the major IT companies (IBM, HP, Dell,

SGI, and Sun) are positioning themselves to associate their names with Grid solutions.

One can observe a clear trend in the initiatives taken by the major IT companies to

promote the outsourcing of Grid Computing, as described in the Clabby Analytics Report

on Grids [5]. However, there are several issues related to the outsourcing of Grid Computing,

such as ownership levels, quality of service, security, liability, systems scalability and com-

patibility, software development etc., that make Grid Computing an interesting case study

for the outsourcing phenomenon.

The basic analysis of the outsourcing scenario provided in this work relies on the assump-

tion that a customer is making the decision to outsource based mostly on the cost-saving

factor. This is an oversimplification of a more complex decision-making process that could

lead to incorrect conclusions.

There appears to be a void in the literature regarding the outsourcing of Grid Comput-

5

Page 7: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

ing. The dearth of coverage on the topic because Grids are a relatively new and complex

technology that researchers in the business field are not entirely familiar with. Furthermore,

Grid Computing has existed primarily in the academic and research world since its inception,

and the technology has not historically been viewed as a resource/service to be outsourced.

Although Grid Computing and IT share some common features, Grid Computing presents

particular challenges that have not been addressed in previous publications and are therefore

worth researching.

The goal of this research is to determine the factors that have the largest influence on

the business decision-making process to outsource Grid Computing.

1.1 Defining Grid Computing

The definitions of Grid Computing used in academia and the business world are, as expected,

almost identical, since the latter is an extension of the former. However, the differences be-

tween the definitions are significant enough to yield the following question: Is a business

Grid a Grid at all?

Ian Foster, considered the father of Grid Computing, defined the characteristics of this

technology as follows [1]:

• Coordinates resources that are not subject to centralized control - (A Grid integrates and

coordinates resources and users that live within different control domains; for example,

the users desktop vs. central computing; different administrative units of the same

company; or different companies; and addresses the issues of security, policy, payment,

membership, and so forth that arise in these settings. Otherwise, we are dealing with

a local management system.)

• Using standard, open, general-purpose protocols and interfaces - (A Grid is built from

multi-purpose protocols and interfaces that address such fundamental issues as authen-

tication, authorization, resource discovery, and resource access. As I discuss further

6

Page 8: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

below, it is important that these protocols and interfaces be standard and open. Other-

wise, we are dealing with an application-specific system.)

• To deliver nontrivial qualities of service - (A Grid allows its constituent resources to be

used in a coordinated fashion to deliver various qualities of service, relating for example

to response time, throughput, availability, and security, and/or co-allocation of multiple

resource types to meet complex user demands, so that the utility of the combined system

is significantly greater than that of the sum of its parts.)”

Academic Grids are based on Foster’s definition of Grid Computing. As such, these

Grids are implemented with decentralized control and emphasis on load balancing (even

distribution of jobs among available resources). Additionally, the academic implementations

are characterized by global scheduling policies that must interact with the policies of the lo-

cal sites, and the complete dissociation between an SU (Service Unit) and its monetary value.

According to IBM, one of the leaders in the Grid market [5], Grid Computing can be

defined as follows [3]: “Grid computing enables the virtualization of distributed computing

and data resources such as processing, network bandwidth and storage capacity to create a

single system image, granting users and applications seamless access to vast IT capabilities.

Just as an Internet user views a unified instance of content via the Web, a grid user essen-

tially sees a single, large virtual computer. At its core, grid computing is based on an open

set of standards and protocols - e.g., Open Grid Services Architecture (OGSA) - that enable

communication across heterogeneous, geographically dispersed environments ”

A business Grid is focused on the virtualization of resources within an organization to

help improve productivity, reduce costs, and create flexible and scalable systems. These

goals are attained by implementing centralized control that guarantees a given quality of

service to its users/clients. The ability to guarantee quality of service is a key ingredient for

the success of Grids in an environment where time corresponds to money gained or lost.

7

Page 9: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Strictly speaking, the requirement for non-centralized control of resources in the definition

provided by Foster practically excludes all business implementations from being considered

Grid Computing. Whether business and academic Grids are the same or not is a discussion

that goes beyond the scope of this work. However, in the interest of provoking thought and

discussion, we can surmise what might occur when one attempts to implement a business

conception of grid in an academic environment and vice versa.

1.1.1 Business Grid in Academia

The business approach to Grids will simply not work in academia due to control issues.

Business Grids are centralized in order to guarantee a certain level of quality of service from

which a business can profit. This centralized control, although technically possible, is unlikely

to happen in an academic or research environment, in which researchers and institutions

prefer to keep their local resources under their own control. The closest implementation of

a business Grid in academia is the alliance formed by several institutions to share resources,

such as the TeraGrid [4]. In this arrangement, each institution dedicates part of its resources

to the newborn Grid, and a new entity of control is created. This new virtual organization is

responsible for allocations, scheduling of jobs, scheduling policies, coordination of resources,

etc. However, this approach is still distinct from a business Grid in the sense that the physical

resources are beyond the domain of control of the virtual organization. This situation makes

the Grid less reliable, because a chain is as strong as it weakest link, and there is no way

for the virtual organization to guarantee the security and the operational integrity of the

individual resources.

1.1.2 Academic Grid in Business

There are two main concerns in the business world when it comes to the implementation of

Grid Computing in a given environment. Those concerns are: quality of service and security

[2, 6]. These two issues are precisely the weaknesses of the academic Grids. Most companies

are very reluctant to use shared resources due to the potential for security breaches. In order

8

Page 10: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

to ensure the highest level of security possible, the organization must have complete control

over its resources. This requirement automatically eliminates the possibility of the academic

Grid working in a business environment. Another key reason for centralized control in the

business environment is the quality of service. In business, there is no place or tolerance

for faulty scheduling, missed deadlines, or down time. These situations equate to money

misspent or lost. In this context, it is clear that the notion of centralized control addresses

the above issues efficiently, therefore delivering a system that features minimum down time

which guarantees the timely completion of jobs.

1.2 Outsourcing Overview

1.2.1 Defining IT Outsourcing

The body of literature on the outsourcing of information technology (IT) was launched in

1992 with the Kodak company’s study of the large scale outsourcing initiative [7]. Since

then, the outsourcing phenomenon has been widely studied and subsequently, several defi-

nitions of outsourcing have been put forth [7, 8, 9, 10].

One of the simplest, most concise definitions of outsourcing is the one provided by Lacity

and Hirschheim [11]. They defined IT outsourcing as follows:

“. . . the purchase of a good or service that was previously provided internally”

The only limitation of this definition is that it confines outsourcing to those goods and

services that were previously provided internally. Although succinct, the definition is inade-

quate as applied to Grid Computing. The Lacity and Hirschheim definition does not consider

work performed by third parties as outsourcing if it hadn’t been performed previously in-

house. The problem with this constraint - as it applies to the outsourcing of Grid Computing

- is that most companies, as described in Section 2.2, do not have a Grid implemented prior

to the hiring of a third party.

9

Page 11: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

In order to address the shortcomings of this definition, I propose a modification to the

outsourcing definition provided by Lacity and Hirschheim. In this context, IT outsourcing

can be defined as:

the purchase of a good or service that was either previously provided internally, or could

be provided internally

The definition of IT outsourcing can be further characterized by the level and amount

of services/goods, and by level of ownership [13, 14]. Consequently, IT outsourcing can be

classified into four categories.

• Total Outsourcing: The decision to transfer the equivalent of more than 80% of the

function’s operating budget for assets, leases, staff, and management responsibility to

external providers. This outsourcing option is associated with problems such as lack of

innovation from the supplier, excess fees for services beyond the contract, fixed prices

that exceeded market prices two years into the contract, etc.

• Total Insourcing: The decision to retain management and provision of more than

80% of the function’s operating budget internally after evaluating the services market.

• Selective Outsourcing: This is the most common type of outsourcing, and it is

defined as the decision to source selected functions from external provider(s) while still

providing between 20% and 80% of the function’s operating budget internally.

• Joint Ventures: The supplier and customer create a new company or business unit.

Deals are typically structured so that the customer investor provides personnel, be-

comes the venture’s first major customer, and shares in future profits if the venture

attracts external customers.

Clearly, this categorization of outsourcing does not apply directly to Grid computing,

since Grids are a part of the IT capabilities of organizations. Therefore, I will describe in

Section 3.2 how a similar classification can be used to identify the different alternatives that

businesses have available when outsourcing Grid computing.

10

Page 12: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

1.2.2 Outsourcing Stages

The outsourcing process can be divided into five distinct stages that reflect the different

factors that businesses consider in their outsourcing analyses, as well as the decisions are

typically made when outsourcing [12]. This model is depicted in Figure 1.

Figure 1: Stage model of IS outsourcing - Source: [12]

• Why: In this stage, the advantages and disadvantages of outsourcing are considered.

• What: Different alternatives of outsourcing arrangements are analyzed.

• Which: This stage is based on the decision that the organization makes when com-

paring various sourcing options.

11

Page 13: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• How: This stage deals with the selection of the vendor, the management of contracts,

and relationships.

• Outcomes: The analysis of the consequences of the sourcing decision, and the assess-

ment of success or failure.

In this study, I will focus on what I consider to be the three main stages involved in the

decision-making process to outsource: Why, What, and How. In Section 3, I will describe

how each of these stages may be applied to during the decision-making process to outsource

Grid Computing technology.

2 Market Analysis

The current rapid advances in technology are constantly creating new business opportunities.

However, to take advantage of such opportunities, companies need to be flexible, dynamic,

and highly adaptable to new demands not only from the customer but from the market

as well. An environment with such characteristics can be easily created by employing the

resources available through Grid computing. In this section, I provide a brief overview of

the current solutions offered by major vendors, their strategies, and their positioning in the

market.

2.1 Grid Solutions in the Market

• IBM: IBM has developed the Grid-enabled e-server product line as a first step towards

a full implementation of Grid in all IBM’s products [18]. IBM offers customer support,

as well as the use of their own facilities, to customers that require a larger volume of

resources. IBM focuses on and supports the development of Grid standards, which

guarantees that any solution provided by IBM will be both flexible and interoperable

with other commercial and non-commercial packages.

12

Page 14: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• Dell: Dell has announced the development of the MegaGrid project. MegaGrid is a

collaborative effort with Oracle, Intel, and EMC to develop a standard approach to

building and deploying an enterprise grid infrastructure that can outperform traditional

HPC solutions at a purported lower cost [27].

• Hewlett Packard: HP has deployed the Globus Toolkit on all their servers running

HP-UX, Linux, and Tru64 UNIX. Additionally, HP has developed its HP StorageWorks

Grid, a virtualized, standards-based architecture for making data a shared resource on

the network [16].

• Sun: Sun is offering their own Grid software called “N1 Grid Engine”. The N1 Grid

Engine software is a distributed management product that optimizes utilization of

software and hardware resources. It lacks the policy management and supports only

Solaris and Linux platforms [17].

• Oracle: Oracle has developed the new generation of database software for Grids. The

Oracle Database 10g and Real Application Clusters coordinate the use of large numbers

of servers and storage, acting as one self-managing Grid for the highest quality of service

on low-cost, modular hardware. Additionally, Oracle Enterprise Manager with Oracle

Grid Control provides a single, integrated interface for administering and monitoring

applications and systems in an Oracle Grid.

2.2 Case Studies

Due to the novelty of Grid Computing in the business world, there exists no literature nor

in-depth studies that analyze the effects of Grids in organizations. In this work, I present

a collection of ten case studies from different sources (IT consultants, vendor websites, etc)

that provide a basic description of the business problems encountered by various companies,

and how the implementation of Grids solved those problems. Unfortunately, the quality of

the case studies does not allow for an in depth analysis of political, cultural, and contractual

factors that might affect the outsourcing of Grids. However, these cases prove very helpful

13

Page 15: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

in understanding companies’ views of the expected savings, the different types of customer-

vendor relationships, and the level of outsourcing being performed in the industry.

Case 1: Grid Technology User Case Study: JP Morgan Chase [23]

• Background: JP Morgan Chase, one of the largest financial services companies in the

US, has been working with partners Platform Computing and Egenera to develop a

Grid initiative to address cost and staffing inefficiencies in their current systems. This

initiative started in 2002, when Steven Neiman, head of high-performance computing at

JP Morgan, realized that there were at least eight instances in which applications were

running on dedicated clusters, mostly made up of Sun Solaris servers. Neiman proposed

to consolidate the eight or nine applications around a single scheduling mechanism and

move them to a shared grid farm using common blade hardware and shared distribution

software. This initiative was called the “Compute Backbone” (CBB) initiative, and

was defined as key strategic development by the Vice President of Architecture at JP

Morgan.

• Grid Solution: Key application architects at JP Morgan decided to replace existing

commercial, and other internally developed distributed computing technologies with a

more general distribution mechanism that would suit all needs. However, such a com-

mercial solution was not available at the moment. Therefore, Platform collaborated

with JP Morgan to develop the system. This new system became Platform Symphony

[22], with JP Morgan retaining rights to the application. Symphony is described as

a policy-driven, real-time application execution layer, which includes workload orches-

tration and service provisioning capabilities.

In addition to Symphony, JP Morgan decided to use Linux as the standard operating

system, along with blade servers that allow for greater scalability and flexibility. A

proof of the scalability of the new system is the incredibly fast growth of the CBB. The

Compute Backbone that started with 700 CPUs currently holds approximately 2000

CPUs.

14

Page 16: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• Results: With the introduction of Linux, and the standardization of all software

builds, it became possible to use one build across all the CPUs in the Grid. As a

consequence, the utilization rates of each CPU has increased, and the management

and economy of scale has improved dramatically.

The cost of the system for JP Morgan was $4.5 million, with a return on investment

(ROI) of 133% in the next two years. In the eyes of JP Morgan, this level of ROI

makes the project a self-funding success.

Case 2: Charles Schwab Responds to Market Conditions and Customer Needs

[24]

• Background: The Charles Schwab Corporation is one of the largest financial services

firms in the United States, serving 8 million active accounts with $758 billion in cus-

tomer assets. The company’s strategy from its conception was to provide the best

customer experience possible, while being highly responsive to the customer’s current

and future needs. To achieve this goal, Schwab needed to evolve its IT infrastructure to

create a new foundation for its business that would provide the flexibility and efficient

use of available resources to support responsive, up-to-date information and advise

across multiple channels.

In 2001, Schwab began to investigate the Grid Computing technology as a medium to

leverage unused computing power within the organization. In 2002, Schwab created

a pilot project to test the technology and prove that it could use Grid within its

environment. One Schwab partner, IBM, helped with the codification and porting of

applications to the Grid environment.

• Grid Solution: Schwab has an internal policy to contract multiple vendors for its

IT infrastructure. Consequently, the resources within the organization are completely

heterogeneous. Schwab needed a Grid solution based on standards that could use het-

erogeneous resources without compromising its vendor independence. To achieve this

goal, Schwab partnered with IBM to create a Grid solution based on the Globus Toolkit

15

Page 17: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

[25], with Linux as the standard operating system. The computational power was pro-

vided by the organization’s current hardware configuration, and newly acquired IBM

eServers xSeries 330 servers. Migration of the applications to the Grid environment

was a joint effort between IBM professionals and Schwab’s application developers.

• Results: One of the results of the Grid implementation is that Schwab now possesses

an IT infrastructure that matches its business model. The flexibility, scalability, and

efficiency of the new system allows Schwab to provide near real-time information to

investors. The processing time of an application was reduced from 8 minutes to 15

seconds. as a result, the customer can make decisions while sitting with a Schwab

advisor, instead of having to go back home and wait for a fax or email.

Case 3: Chicago Stock Exchange Improves Operational Agility [26]

• Background: The Chicago Stock Exchange (CHX) is a strong force for competition

in all US markets. It combines the benefits of a regulated auction market with state-

of-the-art trading technology. In 2002, CHX had reached the limit of the capacity of

its database server. The existing legacy system became complicated to manage and

prone to failure. Recovery from hardware malfunction could take up to 3 hours, and

outages of 20 minutes became the norm. The investment market does not have a steady

level of activity; instead, it is characterized by peaks in which the activity increases

dramatically. In this context, CHX needed to implement a solution that would provide

extreme flexibility, reliability, and scalability. CHX’s solution was the implementation

of Grid Computing.

• Grid Solution: CHX replaced the two HP AlphaServer GS 60 Servers with Oracle

Real Application Clusters, and they implemented Oracle 10g, the database product

for Grids. Additionally, Oracle provided its own Grid manager tool called Oracle

Enterprise Manager Grid Control. This application coordinates the activities within

the Grid across heterogeneous resources.

16

Page 18: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• Results: According to CHX executives, the new system has provided improved cus-

tomer service, and cost savings due to the improved overall manageability of the system.

CHX is expecting a return on investment of 171% over five years.

Case 4: Dell Consolidates European Support System [27]

• Background: Dell is a premier provider of information technology solutions. In 2004,

Dell identified a problem in its data repositories used in Europe, Middle East, and

Africa (EMEA) by business decision makers to produce key reports and business in-

telligence. This system, called Eurostar, provides service to 1200 decision makers, and

it consists of four servers, each running a separate copy of the database. The sys-

tem was becoming increasingly inefficient and difficult to manage as a result of the

need to run multiple database instances to provide scalability. This constant catch-up

game between technology and demand produced problems of synchronization between

databases, data inconsistencies, and poor customer service.

• Grid Solution: In order to repair the fractured infrastructure, Dell decided to create a

new flexible and scalable system that would consolidate all databases. In this way, the

environment would be significantly simplified and the synchronization and customer

service issues would be resolved. The new systems features a cluster of Dell PowerEdge

servers running Oracle Database 10g, with Real Application clusters configured to

operate as a single pool of processing power. The entire system is based on Linux Red

Hat Advanced Server, and it is centrally managed through Oracle Enterprise Manager

Grid Control.

• Results: As a result of this consolidation, the new system provides greater flexibility

and increased manageability. The number of trouble tickets submitted per day was

reduced by 75%, and daily reports are produced 50% faster. The total cost of the

system was $3,814,000 and the expected return on investment (ROI) over 5 years is

172%, yielding a net benefit of $3,336,000.

17

Page 19: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Case 5: Butterfly.net: Powering Next-Generation Gaming with On-Demand

Computing [28]

• Background: Butterfly.net is a development studio, on-line publisher, and infras-

tructure provider for massively multiplayer games that connect players on PCs, game

consoles, and mobile devices. Since its foundation in the year 2000, Butterfly.net made

the strategic decision to build an architecture based on the Grid computing model,

using standard protocols and open-source technologies. The goal was to use advanced

monitoring and routing technologies to dynamically distribute game-related processing

over a large base of distributed, low-cost servers. By providing a means to transpar-

ently shift the processing load to idle resources on an on-demand basis, this system

would eliminate the bottlenecks and reliability issues associated with the gaming in-

frastructure of that time.

• Grid Solution: Butterfly.net selected IBM as a partner to provide a comprehensive

solution. This solution covered all areas of the implementation, hardware hosting,

software configuration and database construction. The Grid implemented consists of

two clusters of approximately 50 IBM eServer xSeries running in IBM’s Sterling, VA

and Los Angeles, CA hosting facilities. The specialized game servers and database

servers are fully meshed over high speed fiber optic lines, enabling the transparent

routing of players to different servers in the Grid.

• Results: Hosting the Grid on IBM’s facilities gave Butterfly.net the flexibility to scale

up as needed and only pay for the computing resources needed. The decision to use

IBM’s hosting facilities allowed Butterfly.net to avoid upfront infrastructure costs. The

scalability of the Grid will allow the connection of over one million simultaneous players

with no compromise in performance. The financial impact of the system is very signifi-

cant as well. A game deployed using the traditional centralized server model generates

profits of $1.6 million, while the same game delivered over grid-based infrastructure

generate profits of $12.8 million.

18

Page 20: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Case 6: Ohio Savings Bank Adopts Enterprise Grid Computing [30]

• Background: The Ohio Savings bank is one of the nation’s largest mortgage lenders

with over 70 branches and nearly $13 billion in assets. In 2004, Ohio Savings started

to display the signs of a rapidly growing business that was significantly stressing its

IT infrastructure. The IT infrastructure suffered from lack of flexibility and scalability

and as a consequence of its overall poor performance, became a drain on financial and

human resources.

The IT executives of Ohio Savings Bank started to consider the implementation of a

Grid in order to address the performance, flexibility, and scalability issues. With this

improved system performance, the bank would be able to lower costs, raise service

levels, and heighten visibility of the bank’s mortgage products.

• Grid Solution: The Ohio Savings bank chose Oracle to provide a comprehensive so-

lution to its infrastructure problem. The IT executives wanted a solution that would

create a service-based architecture for its call center, mortgage, and data warehouse ap-

plications. Oracle provided their suite of Grid applications to coordinate large number

of servers and storage so they could operate like one large computer.

The operating system selected was Red Hat Linux, running Oracle 10g Grid technology.

This new configuration allows the consolidation of servers from multiple vendors into

one common pool of resources.

• Results: The main sources of cost savings in this case comes from the reduction in

personnel needed to support the system as a result of the simplicity of the new environ-

ment, the increased availability of resources, and the reduction of costs of maintenance.

Over five years, Ohio Savings’ investment will yield a return on investment (ROI) of

165%, for a total benefit of $847,000.

19

Page 21: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Case 7: Leading Financial Services Firm Expands Resources with Low-TCO

IBM Grid Solution [31]

• Background: This US-based company (name not released) is a growth-oriented orga-

nization with more than 10,000 employees and 20,000 agents, with revenues of approx-

imately $15 billion. The company offers a broad-based portfolio of financial products

and services, including mutual funds, money management, trust services, life insurance,

and long-term care insurance among others.

During the year 2003, the company’s Unix cluster started to run out of processing

capacity to accommodate the increasing demand for simulations and market forecasts.

In this context, it was imperative that this organization increase the performance and

capacity of its processing infrastructure to allow them to run more simulations and

provide more accurate forecasts. In addition to an increase in computational power,

the organization also required an increase in reliability and fault tolerance.

• Grid Solution: The solution chosen by this company was a Grid implementation by

IBM using IBM eServer BladeCenter running Red Hat Linux. This system was chosen

for its ability to scale out without changing underlying server and operating systems

technologies. Additionally, this new solution consumes less power, takes up less room,

and is easier to maintain than the addition of the equivalent number of machines to

the existing cluster.

• Results: A total cost of ownership revealed that the new solution provided 60%

savings over the traditional alternative of buying more hardware to expand the old

system. In addition to the direct costs benefits, the new system produced a speed-up

of over 50% for current applications. Finally, as a result of the Grid infrastructure

deployed, the system posses greater resilience and disaster recovery capabilities while

providing optimal flexibility and scalability.

20

Page 22: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Case 8: Dell Moves European Sales Operation to Grid [29]

• Background: Dell is one of the largest information technology providers in the US.

It operates one of the highest volume Internet commerce sites in the world. By 2004,

the company was experiencing scalability issues in its Europe-based order management

system, affecting system availability and, potentially, line manufacturing. Additionally,

Dell’s growing volume of customers was threatening to collapse the system that at the

moment could handle a maximum of 6000 concurrent users.

The need to redesign the current system coincided with the company’s new IT in-

frastructure strategy, prompted by Dell CIO Randy Mott, which called for replacing

expensive, slower, proprietary IT technology with Dell’s own standards-based technol-

ogy.

• Grid Solution: Dell wanted a more stable, flexible system with faster performance,

low-cost scalability on demand, and fewer database management tasks. Dell’s IT team

decided to use Oracle’s solution for database management for Grids. The new system,

called GEDIS, runs on Dell PowerEdge 6650 servers with Red Hat Linux. Oracle

10g Grid technology provided database management and consolidation, and cluster

management (load balancing, storage management).

• Results: The new system performed more work more quickly than in the previous

environment. The session capacity has been doubled to 12,000 concurrent sessions.

Additionally, the system automatically distributes and balances the workload in case

of failure of one or more nodes. This disaster recovery technique has dramatically

reduced the possibility of systems outages that could cause gaps in line manufacturing.

The projected return on investment (ROI) over five years is 173%, which represents

total benefits in the amount of $4.6 million. The original investment paid itself within

the first year as result of the lowered operating costs.

21

Page 23: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Case 9: Caprion Pharmaceuticals [32]

• Background: Caprion Pharmaceuticals is a Montreal-based biotechnology company

that applies proteomics (the study of the nature and function of proteins) to discover

and develop pharmaceutical products that enhance the quality of life. Drug discovery

and biological research require vast compute power and massive storage capacity.

IT executives at Caprion viewed the establishment of a massive protein identification

compute farm as an absolute necessity for the business. High performance, reliability,

flexibility, and scalability were the major goals for the new system.

• Grid Solution: Caprion decided to implement a solution provided by multiple vendors

that would allow them to focus on their research instead of Grid deployment. The

vendors selected to form the partnership with Caprion were Sun, Oracle and CGI.

Sun developed the compute farm, providing hardware, the Grid software (Sun Grid

Engine), and the configuration of the system.

In addition, Sun instructed Caprion’s system administrators on all system hardware

and software management. The data warehouse and a closely related data mining

subsystem were developed by the alliance of Oracle and CGI. The data warehouse was

optimized for making queries that call for extensive data searches and analysis, while

the data mining subsystem became the user interface to the data warehouse.

• Results: The implementation of this Grid environment allowed considerable savings

in hardware, and simplified the overall management of the system. Additionally, scala-

bility is no longer an issue of concern, and resiliency and fault tolerance have improved

dramatically.

Case 10: Vanderbilt University Consolidates its Medical Center and University

Onto an Enterprise Grid [33]

• Background: Vanderbilt University is a well-respected research university composed

of ten schools and a teaching hospital. The systems in place at the University and

22

Page 24: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Medical Center were different legacy systems that were running in heterogeneous en-

vironments. Each of the systems had its share of problems: the university servers

couldn’t scale to keep up with data requirements (storage requirements were growing

1000% a year), and the medical center databases had become increasingly expensive

to maintain.

Vanderbilt’s information system executives couldn’t afford to pay for technology in-

crementally in relation to the data growth. In their minds, the solution existed in the

implementation of a comprehensive solution that would consolidate the university and

the medical center systems into one computing environment.

• Grid Solution: In order to realize the consolidation of systems, Vanderbilt Univer-

sity sought the services of Oracle. The Oracle enterprise Grid environment deployed

includes Oracle Application Server and Oracle Real Application Clusters configured

on 16 HP ProLiant DL580 servers running Red Hat Linux.

The system implemented automatically optimized storage performance and eliminated

the need to manually manage data load balancing. This feature resulted in additional

time savings in database creation and management of disk space.

• Results: The new infrastructure resulted in a reduction of costs of approximately 50%

with increased performance. The Grid environment provided Vanderbilt University a

projected benefit of $6.2 million over five years. This amount represents a return on

investment (ROI) of 120%. A key component of the savings experienced by Vanderbilt

University is the reduction in personnel that came as a result of the simplicity of the

new system. This reduction in the need for personnel represents a savings of almost

$2 million over five years. Additionally, the new system could be managed with more

consistency, eliminating the variability caused by manually writing SQL scripts for

maintenance tasks.

23

Page 25: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

3 Outsourcing Grids

As mentioned in Section 1, in this study, I will focus on what I consider the three main

stages, or questions to be answered, involved in the decision-making process to outsource:

Why, What, and How [12]. In the following sections I will describe the issues associated with

each stage applied to Grid computing. The list of factors affecting each stage is not intended

to be exhaustive. However, the list represents the key factors associated with the outsourcing

phenomenon in each stage. The analysis of such factors will be performed based on what can

be inferred from the results of the case studies described in the preceding section, articles in

Grid publications such as GridToday, and interviews with IBM consultants.

3.1 Why Outsource

The question of why companies outsource their IS capabilities has been studied in detail

since Loh and Venkatraman [7] analyzed the Kodak effect, so named because Kodak was

one of the first big companies to outsource a considerable part of its IT department. Loh

and Venkatraman sought to understand the motivation for the sourcing decision. This stage

of the decision making process has been investigated using different lenses that have yielded

different reasons as to why organizations outsource IS. These different research perspectives

vary from innovation diffusion, transactions cost theory, and agency theory, among others

[12].

Additionally, the culture of the organization can play a role in the sourcing decision. A

buyer with a culture that emphasizes financial efficiency is probably more inclined to out-

source that one with a culture that emphasizes stability. A large firm can probably provide

for its own IT services and consider providing them for other firms, but it can afford not to

take advantage of such opportunity. In this context the culture of the organization regarding

innovation will play an important role in the sourcing decision.

The treatment of the different theoretical interpretations of outsourcing is beyond the

24

Page 26: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

scope of this work; instead, I will focus on the four main reasons for outsourcing of IT

identified by Smith et. al [34]. These factors or drivers for outsourcing are: Cost Savings, IS

capabilities or Technical Reasons, Focus on Core Competencies and Environmental Factors.

The definition of each factor and its application to the outsourcing of Grids is presented

below.

3.1.1 Cost Savings

The cost savings factor is often identified as one of the main reasons for outsourcing of IS

[38, 39, 7, 12]. There is a common perception that an outside vendor can provide the same

or a superior level of service at a lower cost than the internal IS department [34]. The logic

behind this belief is that the vendors typically have better economies of scale, more focused

expertise, and increased access to a broader, lower-cost labor pool than the average organi-

zation. This view is particularly appealing in the high performance computing field [21, 41],

where major vendors can provide services and resources at costs that are extremely difficult

to compete with from the internal IS department perspective.

In order to illustrate the costs related to the ownership of high performance resources, I

will provide a simplified version of the total cost of ownership (TCO) for different resources

of different sizes. This sort of analysis would ideally be performed by a company when ana-

lyzing the pros and cons of outsourcing. As shown in the graphs below, the upfront cost of a

new cluster and the subsequent costs of maintenance can be quite large. Depending on the

company’s capital, the cost of new resources may vary depending on whether the company

needs to purchase new hardware, take advantage of unused hardware with a Grid, or buy

processing time in a vendor’s facility.

Additionally, I will provide data provided by some of the case studies summarized in

Section 2.2, where I will show the average return on investment (ROI) that companies have

experienced with Grids, and a detailed breakdown of the percentages of savings associated

25

Page 27: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

with different factors such as hardware, software, labor, etc.

Cost Analysis of HPC resources

The total cost of ownership (TCO) in this work has been calculated in a simplified manner,

taking into account only purchasing costs of hardware, maintenance costs, personnel, and

power requirements for the systems and cooling.

Several assumptions were made in order to obtain monetary values for each factor. The

assumptions made in this study are:

• The power consumption of the systems 65% of the peak consumption.

• Power consumption of the cooling system is equivalent to 40% of the consumption of

the system.

• The number of personnel per system is calculated based on one person per 256 nodes

(this relationship can vary depending on the quality of service that a given organization

wants to provide).

• The salary of system administrators was rounded up to $100,000 a year (including

benefits).

• The price of a kWh was fixed at $0.20. This is a variable quantity that varies according

to the time of the year. However, this value was a rough estimate given by Entergy

personnel with regard to the average amount charged to Louisiana State University

(including discount).

It is worth mentioning other factors that have a considerable effect on the total cost of

ownership, such as the cost of down-time and software costs. However, these costs were not

included in this study for the sake of simplicity.

26

Page 28: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

The example resources selected were SuperMike and Tezpur, which are both supercom-

puters housed at Louisiana State University, IBM P5 575 Server hosted by the Louisiana

Optical Networking Initiative (LONI), and the Funes cluster housed at The University of

Texas at Brownsville (UTB). The criteria for selection was the wide variety of configurations

that these resources offer, as well as accessibility to information on these configurations.

SuperMike - LSU

SuperMike is a 6.267 TFlops Peak Performance, 512 node, dual processor Red Hat Enter-

prise Linux (RHEL) v3 cluster from Atipa Technologies with 3.06 GHz Intel Pentium IV

Xeon processors and 2 GB RAM per node. The original cost of purchase in 2002 was $2

million. In 2004, a system-wide upgrade was performed for a cost of $1 million.

Figure 2: Detail of Costs for SuperMike: the power consumption was assumed to be 100kWh

Tezpur - LSU

Tezpur was designed to replace SuperMike as the top performing cluster in LSU. Tezpur is

a 15.322 TFlops Peak Performance, 360 node, 2 Dual-Core processor Red Hat Enterprise

Linux (RHEL) v4 cluster from Dell with 2.66 GHz Intel Xeon 64bit processors and 4 GB

RAM per node. The cost ofTezpur was one third of SuperMike’s total cost, while offering

almost three times the performance.

27

Page 29: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Figure 3: Percentages of the cost of ownership per category

Figure 4: Detail of Cost for SuperMike: the power consumption was assumed to be 80kWh

IBM Power5 575 Servers - LONI

LONI is in the process of deploying 5 IBM Power5 575 Servers in five different institutions of

higher education across Louisiana. The P5 systems are composed of 14 nodes with 8 2.2GHz

28

Page 30: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Figure 5: Percentages of the cost of ownership per category

Power5 processors and 16 GB of RAM per nodes. The peak performance of this system is

estimated at 0.85 TFlops. The purchasing cost of each P5 is approximately $400,000.

Figure 6: Detail of Costs for IBM P5: the power consumption was assumed to be 35kWh

Funes - UTB

Funes was built at the University of Texas at Brownsville in 2004. It is composed of 64

nodes with dual Pentium Xeon 3.2 Ghz processors, 8 Gigabytes of RAM per node, 2 x 120

29

Page 31: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Figure 7: Percentages of the cost of ownership per category

Gigabyte hard drives per node, and it is interconnected through a gigabit network. The cost

of purchase of this system was approximately $300,000.

Figure 8: Detail of Costs for SuperMike: the power consumption was assumed to be 30kWh

Aggregate Benefits of Grid Implementations

In Section 2.2, I described different case studies on the implementation of Grid computing

in business. Some of these cases, such as cases 1, 3, 2, 6, 8, and 10, provided details on

30

Page 32: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Figure 9: Percentages of the cost of ownership per category

the financial aspects, and the cost savings observed by different businesses when Grids were

implemented. The average return on investment (ROI) experienced in these cases was 155%.

The ROI experienced in these cases is not only a result of increased performance, but is also

a consequence of the avoidance of additional expenses associated with upgrading or replacing

IT, such as the purchase of new hardware and software, or the hiring of more personnel. The

average percentage of savings for these cases are depicted in Figure 10.

The definitions of the categories utilized in Figure 10 are as follows:

• Hardware: The savings experienced in this category reflect the avoidance of purchas-

ing a new system when Grid is implemented.

• Software: In this category, the savings shown represent the number of software licenses

that the company was able to reuse as a result of a Grid implementation.

• Personnel: One of the advantages of Grids is that they have the potential to simplify

an environment. In this case, the headcount needed to operate and maintain the system

31

Page 33: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Figure 10: Aggregate Percentage of Savings per Category, for cases 1, 3, 2, 6, 8, and 10

is reduced.

• Others: This category includes factors such as maintenance costs, housing, etc.

While is is not advisable to generalize these results to every Grid implementation because

of the size of the sample used in this study, and the possible source of bias introduced by

the vendor-supplied cases, the results are still significant enough to be relevant in a cost

analysis. I believe that the saving categories will remain constant across implementations,

and although results and percentages will vary, the analysis performed will remain basically

the same.

3.1.2 Technical Reasons

The technical reasons that might affect the outsourcing decision are directly related to the

internal IT capabilities of an organization. IT departments within firms may struggle to

adjust to rapid changes in technology due to lack of expertise or lack of required equipment

[34]. This expertise, or lack thereof, can work as a trigger for outsourcing, as is the case

32

Page 34: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

when outsourcing is implemented after a failure from the Internal IT department to deliver

a product or service [40].

In the case of Grid computing, the level of internal expertise and resources available may

be key factors that affect the sourcing decision. Considering the words of Walter Stewart

(SGI’s global coordinator of Grid strategy), when he said, “. . . Grids are built not bought”

[41], one can understand the issues that internal IT departments encounter when faced with

this technology. Grid computing is a relatively new technology that is still evolving, and it is

not a simple technology to implement. Since Grids cannot be purchased out of the box, IT

executives must decide whether to allocate time and human resources for an in-house Grid

implementation, or consider the outsourcing option.

The specific technical characteristics of a Grid implementation in business will vary ac-

cording to its purpose. However, the main factors that one should take into account when

considering the deployment a Grid should remain constant. The list of factors includes,

among others: performance, scalability, flexibility, compatibility, manageability, security,

software development, quality of service, and liability. Most of these factors are included

in the case study analysis, while others are drawn from expert opinions [21, 41, 2]. A brief

description of each factors follows:

• Performance: One of the main concerns when implementing a Grid is the overall

performance of the new system. The performance of a Grid system should be superior

to the simple addition of the performance of the individual parts that comprise it.

This characteristic is basically described in point three of Foster’s definition of Grid

Computing, when he requires Grids to provide a non-trivial quality of service [1].

• Scalability: One of the strengths of Grid computing is that it provides an extremely

scalable environment that can accommodate changes in computation or storage needs.

However, the scalability advantage provided by Grid computing depends largely on

proper infrastructure, software implementation, and configuration. It is this need for

33

Page 35: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

configuration that makes scalability a non-trivial state to achieve.

• Flexibility: Another appealing feature of Grid computing related to scalability is

its capacity to provide increased computational power on demand. Different industries

have different needs for computational power, and some industries - such as the financial

market - might experience peaks in demand, when the requirement for computational

power may increase. This characteristic of allocating more resources on demand is

called flexibility. Flexibility might be achieved with the implementation of different

policies for the availability of resources within a Grid, or by the use of Grid facilities

provided by vendors.

• Compatibility: Grid computing is designed to work across heterogeneous environ-

ments. However, a conscious effort should be made on the application development

side to make sure that the application to be run on the Grid will actually run, and

more importantly, run efficiently. In most cases, organizations will have to deal with

legacy applications that were not designed to run on a Grid or a parallel environment.

The development or migration of such applications to work in multiple environments

is a very time-consuming undertaking that might affect the decision to outsource. For

example, in case study 2, IBM provided the migration of the Schwab’s applications to

the Grid environment.

• Manageability: An important factor that was common to almost all case studies

presented was the need for a more manageable system. A Grid implementation should

simplify the system, and make the jobs of system administrators easier. This factor

depends on implementation, and the skill level of the IT department, and how confident

they are with Grid technology.

• Security: This is a very important issue because it is the Achilles’ heel of Grid com-

puting [2]. A Grid implementation must guarantee levels of privacy and security for the

data stored and simulations performed. Grids are, by definition, shared environments,

and different departments or business units within organizations might have different

34

Page 36: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

security policies. This situation may make system administrators uneasy, unless there

is an organization wide policy for security. In the case of Grids provided by vendors,

the problem might arise with the possibility of sharing a computing resource with a

competitor.

• Software Development: When software is developed, there is platform in mind on

which the software will run. In the case of a company outsourcing Grid computing, the

client’s developers might find themselves creating software that would run on a different

platform than the one used in their company. More importantly, the company may

have no control over this platform. Additionally, the internal IT department might

have expertise with a specific environment, thus reducing the options and possibilities

for a Grid implementation.

In addition to the previously described technical issues, there are other issues that arise

with Grid computing, such as Quality of Service (QoS) and Liability.

• Quality of Service: There is a common saying that states, “. . . in business, time is

money.” In this context, obtaining the results of a simulation on time is as important

as getting accurate results. It is key that service providers (internal or external) meet

their promise to complete a job by a certain deadline. Furthermore, the client should

be able to increase the priorities of its jobs to meet unexpected increases in demand.

Obviously, the commitment to meet deadlines should be taken into account in the

price paid for the service (in the case of an external vendor). In addition to policies

that guarantee the completion of a run by a given deadline, QoS also depends on the

resiliency and failure tolerance of the system.

• Liability: There must be a clear definition of the responsibilities regarding the comple-

tion of a simulation. Every single source of failure needs to be taken into account, and

based on that assessment, responsibilities can be determined in case of a failure in the

simulation. This situation applies to in-house developed Grids, but more importantly,

to Grids provided by external vendors.

35

Page 37: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

3.1.3 Core Competencies Reasons

A reason that some companies outsource their IT services is to simplify management and to

focus on the organization’s core business [39]. However, if senior executives see a strategic

potential or advantage within the IT department, they are more likely not to outsource. In

this context, it is key to determine what role Grid computing will play in the overall strategy

of the business.

There are two dominant theories to help managers assess core capabilities to keep in

house. One theory is Transaction Cost Economics (TCE). The other dominant theory is the

Resource-based View (RBV)[14, 12, 35].

• Transaction Cost Economics: TCE addresses decisions of the type “make or buy,”

which makes it a good theory in analyzing sourcing options. The TCE analysis is based

on generic attributes of assets, and describes appropriate ways to govern customer-

supplier relationships [14]. In this context, IT with a high level of asset specificity

(customization), and recurrence should be kept in house, while the rest would be more

efficiently managed if outsourced [42].

It is worth developing the concept of assets specificity due to its major role in TCE, and

its possible impact on the outsourcing of Grid computing. Asset specificity can refer to

relatively unique technical skills, substantial investments in custom-tailored equipment

by external service providers, and extensive business knowledge that is specific to the

environment [43]. Generally speaking, when asset specificity was used to refer to

human aspects, the proposed relationship between high-specificity and the decision to

outsource was supported. However, when asset specificity referred to physical assets,

no significant relation was found with the sourcing decision [12].

• Resource-based View: RBV deals with the production, protection, and acquisi-

tion of resources [35]. This theory suggest that managers will tend to keep valuable,

rare, non-imitable, and non-substitutable strategic assets in-house, while potentially

36

Page 38: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

outsourcing resources consider to be non-strategic [14].

If an organization possesses common and valuable capabilities, they achieve competitive

parity. On the other hand, if the capabilities are also difficult to imitate and are rare,

the organization can obtain and sustain competitive advantage [35]. Finally, RBV

proposes that an IT activity should be outsourced only if it is not a core competency

of the organization [40].

3.1.4 Environmental Factors

Organizations do not exist and perform business in isolation from the rest of the world.

There are factors not specific to the firm, but that exist in its industry or in the economy,

that may affect the sourcing decision [34, 35, 39]. For example, employees of an organization

may believe that IT outsourcing threatens their job security, and thus may resist the sourcing

decision, while stakeholders may believe that IT outsourcing will increase the profitability

of the organization, leading them to favor the decision to outsource.

Another environmental factor that affects the sourcing decision is the tendency of orga-

nizations to imitate the behavior of successful firms [7, 34]. This behavior is often explain

through Innovation Diffusion Theory.

3.1.5 Innovation Diffusion Theory (IDT)

Innovation diffusion theory (IDT) is grounded in sociology and has been used since the 1960s

to study a variety of innovations, such as non-technology tools and organizational innovation.

Everett Rogers wrote the book Diffusion of Innovation [44, 45] in 1962, wherein he defined

diffusion as the process by which innovation is communicated through certain channels over

time by the members of a social system. Rogers focused on perceived characteristics and

how they affect the rate of diffusion. He differentiated adoption from diffusion as a process

that occurs in society, rather than at the individual level.

37

Page 39: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Rogers developed a curve to depict the adoption of innovation in business. This curve is

depicted in Figure 11.

Figure 11: Rogers’ Innovation Adoption Curve

A brief description of the characteristics of the type individuals that personify each stage

is as follows:

• Innovators: Brave people that pull the change effort. Innovators are very important

communication mechanisms.

• Early Adopters: Respectable people that are often opinion leaders. They tend to

try out new ideas early on, but in a cautious way.

• Early Majority: Thoughtful people that are careful but accept change more quickly

than average people do.

• Late Majority: Skeptical people that will adopt new ideas or products only when

the majority is using it.

38

Page 40: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• Laggards: Traditional people that love to stick to the “old ways” and are critical about

new ideas. They will only accept new ideas if the new ideas have become mainstream

or even tradition.

Several analysts agree that Grid computing is gaining broader acceptance in the business

environment [21, 5]. This agreement implies that Grid computing as an innovation is moving

from the Early Adopters to the Early Majority Stage. This shift to the third stage implies the

potential for a “Grid Boom”, where Grid Computing could definitely become a mainstream

technology in the business world.

3.2 What Needs to be Outsourced

Once an organization has decided that outsourcing is a beneficial move, the question that

arises next is what activity or function needs to be outsourced. According to Hirschheim

et al. [12] this question can only be answered if two conditions are fulfilled. First, at least

two different options have to be available. Second, there needs to be a reason or a rationale

that serves as a selection criterion. This last requirement is related to the reasons behind

the outsourcing decision (why outsource).

One of the determinants to answer the question of “What needs to be outsources” is the

operational definition of the IT function in an organization. If the IT function of an organi-

zation is thought of as a portfolio of different functions that are subject to outsourcing, the

important question is how those functions are defined. If the functions are believed to be

independent of each other, the sourcing of one of them might overlook the importance of an

interdependency with other functions. This may lead to the failure of the sourcing scenario.

On the other hand, if all functions are considered to be interrelated, the sourcing decision

becomes more difficult, and it could impact not only the sourcing decision, but also the type

of outsourcing arrangement selected (total, selective, joint venture) [12].

39

Page 41: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Another factor that determines what will be outsourced is the type of sourcing arrange-

ment available to the organization. In Section [?], four categories describing different levels

of outsourcing were presented: Total Outsourcing, Total Insourcing, Selective Outsourcing,

and Joint Ventures). However, these definitions do not apply to the Grid computing envi-

ronment, since they refer to the outsourcing of the IT function of an organization as a whole.

In this section, I will redefine these categories in the context of Grid computing.

3.2.1 Total Outsourcing

Total Outsourcing can be defined as the decision to transfer the responsibilities for the im-

plementation, configuration, management, maintenance, and housing of resources to a third

party. In this sort of arrangement, the client essentially buys a service that allows the com-

pany to use a Grid built in the vendor’s facility.

An example of this type of outsourcing deal is case study number 6, “Butterfly.net: Pow-

ering Next-Generation Gaming with On-Demand Computing.” In this case, IBM provided

Butterfly.net with a comprehensive solution covering all areas of the implementation, hard-

ware hosting, software configuration and database construction.

One of the most important advantages of this arrangement is the avoidance of large

upfront hardware investments. This type of outsourcing allows for a maximum level of scal-

ability and flexibility, since the client can always adjust its demand for resources according

to the level of activity of the company. Additionally, total outsourcing does not require the

client’s IT department o become Grid experts, and they can focus on their normal activities.

The most obvious disadvantage of this type of outsourcing is the dependency relationship

that is created with the vendor. Moreover, in this type of sourcing deal, contract management

and client-vendor interaction become essential factors for success. These factors are impor-

40

Page 42: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

tant because the performance of the vendor and its quality of service is directly proportional

to the performance of the client and the quality of service that can be provided.

3.2.2 Total Insourcing

Total Insourcing is the decision to internally manage the entire Grid building process, from

implementation, management, configuration and maintenance, to the housing of the re-

sources. This type of arrangement is not very common in the business world, but is more

often found in academia.

An example of such a sourcing decision is the Grid built by the Center for Computation

and Technology (CCT) at Louisiana State University.

3.2.3 Selective Outsourcing

Selective Outsourcing occurs when the Grid is built with the “collaboration of a third party.”

In this arrangement, the client would typically house the resources (pre-existing or new), and

the vendor would provide expertise in consolidating its resources with Grid middleware (this

can be proprietary software or open source). The level of involvement of the vendor will vary

on a per case basis; however, the control and ownership of the Grid always lies with the client.

This is the most common type of sourcing arrangement. It usually leverages pre-existing

relationships with vendors to make a more efficient use of available resources. By leveraging

these pre-existing relationships, clients make the client-vendor interaction stronger, they have

more experience in contract management, and they are more likely to experience savings in

areas such as software licensing and training.

3.2.4 Joint Ventures

Join Ventures, or Strategic Alliances, take place when the client and vendor share the respon-

sibility for the development of a system. This mean that both client and vendor share the

41

Page 43: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

risks, failures, and profits of the Grid implementation. This is the distinctive characteristic

of Joint Ventures, since in no other arrangement does the vendor or service provider risk

experiencing a loss.

Although this type of arrangement is not very common, it is found often enough to be

considered a real possibility, more so when the client posses a novel application or resource

that complements the services offered by the vendor.

Examples of this type of outsourcing can be found in case study number 1, “Grid Tech-

nology User Case Study: JP Morgan Chase”, where Platform collaborated with JP Morgan

on the implementation of its Computer Backbone system. This new system became Platform

Symphony [22], with JP Morgan retaining rights to the application. Symphony is described

as a policy-driven, real-time application execution layer, including workload orchestration

and service provisioning capabilities [23]. Another example of a Joint Venture is the partner-

ship developed between Schlumberger and IBM to performed reservoir uncertainty analysis

using a Grid enabled version of Schlumberger’s Cougar software [46].

3.3 How to Outsource

Once the decision to outsource has been made, and the organization has selected the ar-

rangement of sourcing that is most convenient and beneficial to them, the next logical step

is to decide how to carry out the outsourcing process. This stage deals mainly with three

actions: vendor selection, contract negotiation, and contract/relationship management [12].

3.3.1 Vendor Selection

There are multiple arrangements that can describe this relationship, such as one client -

one vendor, one client - many vendors, many clients - many vendors, many clients - many

vendors. In this study I will only focus on the two most common relationships, one client -

one vendor, and one client - many vendor [37].

42

Page 44: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• One Client - One Vendor: this is the most common relationship studied in the

outsourcing research. The client relies on a single vendor to satisfy all of its outsourcing

needs. These types of deals often involve a major vendor that is equipped with the

market power and knowledge expertise to provide a comprehensive IT solution [37].

An advantage of this type of relationship is the potential for cost savings as a result of

having one vendor providing a large set of services. On the other hand, having only one

vendor may lead to proprietary solutions that can make the adoption of other products

very difficult.

Examples of these kind of arrangements are provided in several of the case studies

described in Section 2.2. As an example, Case 4: “ Dell Consolidates European Support

System [27],” shows how Dell (functioning as a client) obtains the services of Oracle

to upgrade its database system to a Grid Solution.

• One Client - Many Vendors: In this scenario, one client utilizes multiple vendors

to achieve the its goals. The division of labor and responsibilities among vendors is

jointly negotiated and understood by all parties [37].

The advantage of this type of arrangement is the ability to fit every need with a vendor

whose strength is in the same area, thereby obtaining optimal results. However, these

many vendor deals might can be extremely difficult to handle at the contract level and

management level.

An example of a One Client - Many Vendors deal was described in case study number

9: “Caprion Pharmaceuticals [32]”, wherein Caprion selected Sun, Oracle and CGI to

develop their system. Sun developed the compute farm, providing hardware, the Grid

software (Sun Grid Engine), and the configuration of the system. Oracle and CGI

developed the data warehouse and a closely related data mining subsystem that works

as user interface for the data warehouse.

43

Page 45: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

3.3.2 Contract Negotiation

There is agreement in the literature that contract negotiation is an important factor deter-

mining the success of a sourcing experience [35, 14, 12]. During the negotiation stage, the

client and vendor communicate with each other to exchange information about each others’

capabilities, positions and interests. Eventually, both parties will decided whether to estab-

lish a sourcing relationship or not.

Willcocks et al. [14], define four different types of contracts:

• Standard Contracts: The customer signs the supplier’s standard, off-the-shelf con-

tract.

• Detailed Contracts: The contract includes special contractual clauses for service

scope, service levels, measures of performance, and penalties for non-performance.

• Loose Contracts: The contract does not provide comprehensive performance mea-

sures or contingencies, but specifies that the suppliers perform whatever the customer

was doing in the baseline year for the duration of the contract at 10-30% less than the

customer’s budget.

• Mixed Contracts: For the first few years of the contract, requirements are fully

specified. However, long term requirements cannot be defined in a detailed manner.

According to the findings of Willcocks et al. [14], detailed contracts achieved expecta-

tions with greater frequency than the other types. In these cases, organizations understood

the functions to be outsourced fairly well, and they could therefore define their precise re-

quirements in a contract. Additionally, organizations developing detailed contracts spent up

to 18 months negotiating the clauses of the contract. This careful process of negotiation

revealed a 75% rate of successful contracts.

44

Page 46: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

The most common clauses included in detailed contracts, and the frequency of their

occurrence are depicted in Figure 12. In addition to these relatively common clauses, detailed

contracts are also starting to include mechanisms of change, to account for unforeseen changes

in the market, or parties involved. These new features include [14]:

• Planned contract realignment points to adapt the contract every few years.

• Contingency prices for fluctuation in the volume of demand.

• Negotiated price and service level improvements over time.

• External benchmarking of best-of-breed suppliers to reset prices and service levels

Figure 12: Clauses included in detailed contracts

Length of the Contract

Another important feature of the contract, aside from the clauses included, is its duration.

There is support in the literature for the assumption that shorter contracts yield better

45

Page 47: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

results than long-term contracts [36, 35, 12, 14]. We can consider a contract of three years

or less to be a short-term contract.

Willcocks et al. [14], performed an analysis of 85 case studies, in order to assess the

relationship between contract length and success rate. The findings of the study reveal that

87% of sourcing decisions governed by contracts of three years or less were successful. How-

ever, only 38% of sourcing decisions with contracts of eight years or longer were considered

successful.

The reasons why short term contracts generally yield better results can vary. However,

generally , we can say that this type of contract involves less uncertainty, allows participants

to recover from mistake more quickly, motivates supplier performance, and more importantly

it helps to ensure that the price charged for a given service is fair compared to the market

situation.

3.3.3 Contract Management

A good contract is necessary but not sufficient for a successful IT outsourcing project [47].

Since market conditions are bound to change, the contracts may not include the foresight

sufficient to cope with changes. Thus, it is imperative that contract management activities

be performed to potentially lead to renegotiation of the contract [35].

In addition to renegotiation skills, a client should also exercise periodic control over

the fulfillment of the contract. Performance standards and control mechanisms should be

clearly defined. According to McFarlan, it is critical for the customer to maintain the

capabilities that enable it to deal with contractual issues in a constantly evolving technical

and competitive environment [40].

46

Page 48: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

4 Grid on the Bayou

The Louisiana Optical Network Initiative (LONI) is a $40 million effort to create a statewide

optical network connecting 8 research institutions (LSU, Louisiana Tech University, LSU

Medical Centers in Shreveport and New Orleans, the University of Louisiana at Lafayette,

Southern University, University of New Orleans, and Tulane University) across the state of

Louisiana as depicted in Figure 13. Each research institution will host an IBM P5 server,

which will give this network a total computational power of around 5 Teraflops.

Figure 13: LONI Map

LONI will also allow Louisiana to be a member of the National Lambda Rail (NLR),

“a major initiative of U.S. research universities and private sector technology companies to

provide a national scale infrastructure for research and experimentation in networking tech-

nologies and applications” [48]. The access to NLR will be provided by a server located at

LSU in Baton Rouge. This local point of access to such advanced technology represents an

invaluable opportunity for the city of Baton Rouge to develop into an IT-oriented metropolis.

The state will be equipped to be not only to keep the talented people from our universities,

but will attract talent from all over the country and the world.

47

Page 49: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

4.1 Economic Development

Since its inception, LONI has consistently maintained the economic development of the re-

gion as one of its main goals [50]. LONI has the potential to create a statewide computing

environment that will allow companies to flourish in this technologically driven era. Addi-

tionally, research institutions across the state will be able to perform top quality research by

leveraging LONI resources.

4.1.1 A Common View of Economic Development

As mentioned in the LONI Concept paper [50], the presence of an advanced environment for

research and development is in and of itself a crucial factor in the states ability to support

and further develop the intellectual environment needed to foster economic development.

The establishment of a stronger research base in the states universities will be critical in the

state’s efforts to attract not only other strong faculty and students, but also business and in-

dustry. This view is shared by Bob Fudickar, technology industry director for the Louisiana

Department of Economic Development and member of the LONI Management Council:

“The dynamics of academics, research and economic development depend heavily on tech-

nological superiority,” Fudickar said. ”Many of our state’s current and future startup tech-

nology companies stand to benefit significantly from LONI, especially in the area of product

development.”

4.1.2 A Comprehensive View of Economic Development

The above view of economic development is mainly driven by technological infrastructure

and intellectual capital. Although technically correct, it may be an oversimplification of a

48

Page 50: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

much more complex phenomenon.

Intellectual capital as an economic driver is key in the development of any technological

park, such as those in Silicon Valley, or Austin, Texas. However, this is a goal that, depend-

ing on the location and key players, could take several years to materialize, maybe longer

than the life of the LONI project.

In this context, other actions could be taken in order to achieve short-term benefits, while

still striving for the longer term goals, such as the acquisition and development of intellec-

tual capital. Such actions encompass a more aggressive interaction of LONI with businesses.

However, in order for LONI to efficiently interact with businesses in a consistent way, some

changes should be made to its governance structure.

4.1.3 A New Structure to Foster Economic Development

Governor Blanco stated that one way to leverage the state’s investment in LONI is to re-

serve 10% of the grid’s computational resources for its interaction with businesses [51]. The

important question here is not whether 10% is the right amount or not; instead, the question

is: How can LONI reach out and interact with business to achieve the desired goal of job

creation, and other indirect economic impacts?

One possible answer to such a question is the modification of the governance structure of

LONI to shift its complete academic focus to a more balanced business-academia governance.

In this context, an executive management team could be created, and a body of governance

above them to oversee their actions and progress. The governing authority should be formed

by a mix of researchers and industry members.

This new structure would allow LONI to take definitive and tangible actions towards

economic development. Some of these actions include:

49

Page 51: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

• Quantification of Interest: This entails the assessment of the level of interest from

the private sector with regard to utilizing LONI for their business activities, and an

assessment of the potential top players in industry. This quantification could provide

a further understanding of the expectations that different industries have for LONI in

terms of capabilities, services, and support.

• Pool of Knowledge: Based on the previous step, LONI should create a “pool of

knowledge”, in order to effectively interact with the industry. This “pool of knowledge”

would basically consist of a detailed database describing the different areas of research,

projects, and personnel associated with LONI. The implementation of such a knowledge

database could be key to assessing the strengths and weaknesses in the different research

areas. Based on this assessment, LONI’s management would be in a condition to

determine which areas of research need further development, what kind of researchers

should be considered for hiring, what type and level of resources should be allocated

to each area, and more importantly, it would allow LONI management to develop a

comprehensive research agenda.

• Allocation Procedures: This item is intended to describe the procedures that need

to be followed by a business in order to access LONI resources. This includes pricing

schemes, different levels of service and support, allocation policies, and conditions of

use among others. The development of these procedures might bring as a consequence

the creation of a customer support department within LONI. This is a key step that

LONI must take in order to have any meaningful interaction with businesses.

• Proactive Marketing Strategy: Once the previous three steps have been completed,

LONI should create a marketing team with an aggressive strategy to attract businesses.

Such a strategy should be targeted to potential customers (determined in the action-

item), leveraging LONI’s strengths (described in the second action-item), and offering

different levels of service and prices (depicted in the third action-item). In this context,

LONI should work closely with regional economic development offices, such as the

50

Page 52: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

Louisiana Department of Economic Development.

LONI’s legal situation is a very complex one because there are contractual restrictions

that limit LONI’s involvement with businesses. These constraints affect LONI’s potential

economic development impact on the region. A possible solution to this problem could be the

creation of a LONI-supported Technology Incubator Center, where industries could benefit

from LONI’s infrastructure while their products are still in development. Moreover, indus-

tries could be required to collaborate in their R&D efforts with the pertinent researchers,

depending on the area of investigation. This is similar to the academic research requirement

that businesses must adhere to in order to obtain federal SBIR/STTR funding.

A center with these characteristics would increase LONI’s visibility in the industry, and

will further the broader goal of attracting and producing intellectual capital in Louisiana.

5 Conclusion

Grid Computing presents a fascinating scenario for the study of the outsourcing phenomenon.

The technical characteristics of Grids, and the implications of Grid implementations for the

management structure of organizations, present unique opportunities for the research of the

sourcing effect.

Grid Computing is slowly becoming a mainstream technology that has been adopted

across different industries and academia. This adoption trend [5] should be accompanied by

an increase of literature documenting the effects of Grid implementations on businesses, and

the study of the sourcing phenomenon applied to Grids.

There appears to be a void in the literature regarding the outsourcing of Grid Computing.

In this context, this work represents the first step to understanding the factors that have the

largest influence on the business decision-making process to outsource Grid Computing.

51

Page 53: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

References

[1] Ian Foster, ”What is the Grid? A three point checklist”, Grid Today, Vol. 1, NO. 6,

July 22, 2002.

[2] Edmund X. DeJesus, ”Grid computing and security uncertainties”, SearchSecurity.com,

March 30, 2006.

[3] http://www-03.ibm.com/grid/about grid/what is.shtml

[4] http://www.teragrid.org/

[5] www.ibm.com/grid/pdf/Clabby Grid Report 2004 Edition.pdf

[6] http://www.gridcomputingplanet.com/news/article.php/3523061

[7] Loh, L. and Venkatraman, N., “Diffusion of Information Technology Outsourcing Influ-

ence Sources and the Kodak Effect”, Information Systems Research, December 1992,

334-358.

[8] Chaudhury, A., Nam, K. and Rao, H. R., “Management of Information Systems Out-

sourcing: A Bidding Perspective,” Journal of Management Information Systems, Vol.

12, No. 2, pp. 131-159.

[9] Cheon, M. J, Grover, V. and Teng, J. T. C., “Theoretical Perspectives on the Outsourc-

ing of Information Systems,” Journal of Information Technology, Vol. 10, pp. 209-210.

[10] Fitzgerald G. and Willcocks, L. P., “Contracts and Partnerships in the Outsourcing

of IT, ” Proceedings of the 15th International Conference on Information Systems,

Vancouver, Canada, pp. 91-98.

[11] Lacity, M. C. and Hirschheim, R. A., “Information Systems Outsourcing: Myths,

Metaphors, and Realities,” Chichester, New York:Wiley.

52

Page 54: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

[12] Jeans Dibbern, Tim Goles, Rudy Hirschheim, Bandula Jayatilaka“Information Systems

Outsourcing: A Survey and Analysis of the Literature”, ACM SIGMIS Database, Vol-

ume 35 , Issue 4, 2004.

[13] Lacity, M. C. and Hirschheim, R. A. (1995). Beyond the information systems outsourcing

bandwagon : the insourcing response, Chichester ; New York: Wiley.

[14] Leslie Willcocks, Mary Lacity, Sara Cullen, “Information Technology Sourcing: Fifteen

Years of Learning”, Chapter 13, Mansell, R. et al. Handbook, Oxford, 2006.

[15] http://www1.us.dell.com/

[16] http://h71028.www7.hp.com/enterprise/cache/250417-0-0-0-121.html

[17] http://www.sun.com/software/gridware/5.3/index.xml

[18] http://www-1.ibm.com/grid/

[19] http://hst.home.cern.ch/hst/publications/bigresults-final.pdf

[20] http://www.pegasusresearch.org.uk/

[21] http://www.cioupdate.com/reports/article.php/2178091

[22] http://www.platform.com/Products/Platform.Symphony/

[23] The 451 Group, “Grids 2004: From Rocket Science to Business Service”, 2004

[24] David S. Marshak, “Charles Schwab Responds to Market Conditions and Customer

Needs. Servies Oriented Architecture Improves Time to Market and Leverages Existing

Investments.”, December 2003, IBM Corporation

[25] http://www.globus.org/

[26] http://www.oracle.com/customers/studies/roi/chicagostockexchange.pdf

[27] http://www.oracle.com/customers/studies/roi/delleurostar.pdf

53

Page 55: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

[28] www.ibm.com/grid/pdf/butterfly.pdf

[29] http://www.oracle.com/customers/studies/roi/dellgedis.pdf

[30] http://www.oracle.com/customers/studies/roi/ohiosavingsbank.pdf

[31] http://www.ibm.com/grid/pdf/mmblinded.pdf

[32] http://www.sun.com/products/hpc/pdfs/caprion.pdf

[33] http://www.oracle.com/customers/studies/roi/vanderbilt.pdf

[34] Michael Alan Smith, Saby Asachi Mitra, and Sridhar Narasimhan, “ Information Sys-

tems Outsourcing: A Study of Pre-Event Firm Characteristics.” Journal of Management

Information Systems, Fall 1998, Vol. 15, No 2, pp. 61-93.

[35] Pamsy P. Hui, Nils O. Fonstad, Cynthia M. Beath, “Information Technology Service

Sourcing, A Framework for Research,” work in progress.

[36] Currie, W.L., and Willcocks, L.P. “Analysing four types of IT sourcing decisions in

the context of scale, client/supplier interdependency and risk mitigation,” Information

Systems Journal (8) 1998, pp 119-143.

[37] Michael J. Gallivan, Wonseok Oh, “Analyzing IT Outsourcing Relationships as Alliances

among Multiple Clients and Vendors,” Proceedings of the 32nd Hawaii International

Conference on Systems Sciences, 1999.

[38] Alpar, P., and Saharia, A.N. “Outsourcing infomiation system functions: an organiza-

tion economics perspective.” Journal of Organizational Computing, 5, 3 (1995), 197217.

[39] Lacity, M.C; Hirschheim, R.; and Willcocks, L. “Realizing outsourcing expectations.”

Information Systems Management, 11, 4 (Fall 1994), 7-18.

[40] McFarlan, F.W., and Nolan, R.L. “How to manage an IS outsourcing alliance.” Sloan

Management Review, 36,2 (Winter 1995), 9-23.

54

Page 56: Out so urci ng o f Gr id C o m put inggallen/Students/Pena_2007.pdf3 Out so ur cing Gr ids 2 4 ... liability, systems scala bility and co m-pat ibility, so ftw ar e dev elopmen t etc.,

[41] http://www.gridtoday.com/04/0920/103826.html

[42] Aubert, B. A., Rivard, S. and Patry, M. (1996). “A Transaction Cost Approach to

Outsourcing Behavior: Some Empirical Evidence,” Information & Management, Vol.

30, No. 2, pp. 51-64.

[43] Ang, S. and Cummings, L. L. (1997). ”Strategic Response to Institutional Influences on

Information Systems Outsourcing,” Organization Science, Vol. 8, No. 3, pp. 235-256.

[44] Rogers, Everett M., “Diffusion of Innovation,” New York, NY: Free Press, 1962.

[45] Rogers, Everett M., “Diffusion of Innovation, Fifth Edition,” New York, NY: Free Press.

ISBN 0-7432-2209-1, 2003

[46] http://www.slb.com/media/services/software/reseng/ressim ondemand.pdf

[47] Willcocks, L.P., and Kern, T. “IT outsourcing as strategic partnering: The case of the

IK inland Revenue,” European Journal of Information Systems (7) 1998, pp. 29-45

[48] http://www.nlr.net/

[49] https://www.loni.org/

[50] Gabrielle Allen, Charles MacMahon, Ed Seidel, Tom Tierney, “LONI Concept Paper,”

(2003)

[51] http://cenit.latech.edu/cenit/news/index.asp?name=view&article=20060919

55