TNC2011

Terena Prague

Dr Malcolm Read

JISC Executive Secretary

Chair of e-infraNet Steering Committee

18/05/2011 | Slide 1

18 May 2011

Usage

Mostly SaaS to date: e-Mail, followed by storage, web services and virtual learning environments. Some use of other levels of stack

Reason for uptake

Provision of a better service, followed by a reduction in costs, better collaboration and a reduction of hardware overheads as part of a green IT strategy

Barriers to uptake

Jurisdiction issues, uptime of services, security of services, lock-in and the strength of service level agreements

Real costs of research computing currently hidden from most users so Cloud can look expensive

18/05/2011 | Slide 2

Current Picture

Software as a Service (SaaS)– eg, Google Apps, Microsoft 365

Platform as a Service (PaaS)– eg, MS Azure, Google App Engine

Infrastructure as a Service (IaaS)– ie, compute, storage, database

– eg, Amazon Web Services, VmwarevCloud, Eucalyptus

18/052011 | Slide 3

Which bit of the Cloud?

Do we need to add in Middleware as a Service (MaaS)– Offerings like Cassandra or Amazon Elastic Map Reduce

Dia

gram

cre

dit:

ww

w.s

aasb

logs

.com

Every level of the ‘stack’ is catered for – which is right for you?

Efficient, cost effective infrastructure – provides access to industrial scale economies in deployment and

use of infrastructure and applications

– leads to financial benefits and in turn carbon and other environmental benefits

Potential to cope with sudden peak demands for increased storage and compute requirements

Provides a suitable ‘neutral platform’ for HEI / business collaboration.

Lowers barriers to participation in high end computing – “e-Science for the Masses”

05/07/2011 | Slide 4

Why Invest in Cloud Computing?

Large data centres can use economies of scale to be significantly cheaper and can be flexible in delivering layers of a standardised, modularised service

Particularly attractive to smaller institutions without the capital budget for wholesale rip and replace that are able to secure access to upgraded infrastructure which they could not otherwise afford

NB Requires changes in culture – specifically expectations of ‘fine-tuning’ of services to meet specific requirements

For example, the European Space Agency is utilizing Amazon EC2 for the data-processing needs of its Gaia mission. The 40Gb per night that Gaia will generate would have cost $1.5 million using local resources but research suggests it could cost in the $500,000 range using EC2

05/07/2011 | Slide 5

Financial Cost

Having more storage and compute available on demand is very useful for dealing with sudden peak usage and projects that on occasion need to crunch larger data

Examples– Flood simulation at Newcastle University

– Machine Learning Group at the University of Cambridge used Amazon’s Elastic Map Reduce to process data sets that would have taken weeks to do locally

– ApatMEMS-ID developing means of distinguishing different strains of MSRA. Cloud is used as “cloudburst” to supplement local Condor grid

05/07/2011 | Slide 6

Capacity

Using Cloud can make it easier to collaborate with businesses

For instance, if a spin-off company should come out of a project that uses web resources it may be easier to hand over control to something run on a virtual machine

Similarly, for partnerships with industry using an external cloud provider can make it easier to collaborate as both HEIs and industry often restrict external access to their systems

Services like e-Science Central1, a ‘Science-as-a-Service’ platform that combines Software-as-a-Service, social networking, and “cloud” computing remove the need to maintain one’s own systems whilst still offering control over what, when and with whom to share data

1 www.esciencecentral.co.uk05/07/2011 | Slide 7

Collaboration

Good candidate scenarios for Cloud deployment that JISC has identified include those with one or more of the following characteristics:– Short timescale requirements

– Infrequent use and/or no desire to maintain infrastructure

– Dynamic scalability to larger capacity (‘cloudbursting’)

– Transfer to commercial use

– Flexibility with system configuration and/or frozen system configuration

– Data hosting and backup

– Cloud-based research publications

– Ad hoc activities in support of research

05/07/2011 | Slide 8

What should we look to use the Cloud for?

The key word here is trust. Researchers and HE staff need to be able to trust the reliability and integrity of the cloud they use as well as sustainability of data and overall reliability

There are four critical enabling factors of Cloud Computing: – virtualisation and automation

– pay per user software

– data centres

– broadband connectivity

There's no reason though why the research and HE Cloud can't be built from commercial components

However the Cloud itself and the HEI’s connections to it must be as safe and reliable as the physical IT hardware it is replacing otherwise the benefits are meaningless 05/07/2011 | Slide 9

Why build a private Cloud?

NRENs are obvious candidates for brokering and delivering private cloud services, exploiting their procurement and service delivery expertise.

Security (e.g. access management) is also usually managed by the NREN.

05/07/2011 | Slide 10

Role for NRENs

If and when countries build National Research and Education (private) Clouds there is value in exploiting economies of scale at the European level.

Clear parallel with NRENs and GEANT

05/07/2011 | Slide 11

Federation of NRECs?

Using Cloud Computing for Research1 study

Report from workshop on Cloud Computing2

Currently 11 pilot projects funded by JISC and EPSRC3

1 www.jisc.ac.uk/whatwedo/programmes/researchinfrastructure/usingcloudcomp.aspx2 www.rcuk.ac.uk/documents/research/esci/CloudWorkshopJuly2010.pdf3 cloudresearch.jiscinvolve.org

05/07/2011 | Slide 12

Relevant UK studies