Storage Basics

A Guide to the World of Storage Technology

an Storage eBook

Storage Basics:

1

contentsThis content was adapted from EarthWeb'sEnterprise Storage Forum Web site. Contributors:Dan Muse, Paul Shread, Drew Robb, MikeHarwood, and Henry Newman.

Solving Storage for Your SMB, An Internet.com Storage eBook.© 2007, Jupitermedia Corp.

2 IntroductionMichael Pastore

3 What Makes a Storage Server a Storage Server?Drew Robb

6 Storage Strategies Made SimpleDrew Robb

8 Storage Security BasicsDrew Robb

11 Storage Budgeting TipsHenry Newman

3

3

6

118

Storage Basics: A Guide to the World of Storage Technology[ ]

Sales of storage products reached $3.9 billion inthe fourth quarter of 2005, according to IDC, thebest quarter for the storage market since IDC

began tracking it in 2001. You can expect the numbersto keep rising.

Regardless of industry, size, or age, enterprises areawash in more data than ever before. Fewer businessprocesses rely on paper, and the file cabinets that oncefilled offices for generations are now located on racks inthe server room. Federal regu-lations require that data bestored, protected, and retriev-able for a certain amount oftime, and specific industry reg-ulations add to the burden.

Storage is one of the mostbasic operations performed bycomputers, yet it continues toevolve. In the days of main-frames, data was stored physi-cally separate from the actual processing unit, but wasstill only accessible through the processing units. As PC-based servers became more commonplace, storagedevices went “inside the box” or in external boxes thatwere connected directly to the system. Each of theseapproaches was valid in its time, but as our need tostore increasing volumes of data and our need to makeit more accessible grew, other alternatives were needed.

Network storage is a generic term used to describenetwork-based data storage, but there are many tech-nologies within it. Direct Attached Storage (DAS) is astorage device that is directly attached to a host sys-tem. The simplest example of DAS is the internal harddrive of a server computer, though storage deviceshoused in an external box come under this banner aswell. DAS is still, by far, the most common method of

storing data for computer systems.

Network Attached Storage, or NAS, is a data storagemechanism that uses special devices connected directlyto the network media. These devices are assigned an IPaddress and can then be accessed by clients via a serv-er that acts as a gateway to the data, or in some casesallows the device to be accessed directly by the clientswithout an intermediary.

A Storage Area Network(SAN) is a network of stor-age devices that are con-nected to each other and toa server, or cluster ofservers, which act as anaccess point to the SAN. Insome configurations a SANis also connected to the net-work. SANs use specialswitches as a mechanism toconnect the devices. Theseswitches, which look a lot

like a normal Ethernet networking switch, act as theconnectivity point

Why is it important to learn the basics of storage tech-nology? As mentioned earlier, how enterprises storedata is becoming more than a best practice, it'sbecoming a legal matter as well, and the penalties forindividuals and corporations can be severe.

Storage is also a growing area within IT, which meansemployment opportunities exist now, and should existfor some time. According to a one study, fewer than 25percent of either Unix-/Linux- or Windows-based ITorganizations had their own storage management teamat the end of 2004. By the end of 2006, however, thatnumber is expected to soar above 75 percent. ■

2 An Internet.com Storage eBook. © 2007, Jupitermedia Corp.


Storage Basics:A Guide to the World of Storage Technology

Introduction by Michael Pastore

Jupiterimages

Ask people what a storage server is, and you canexpect to hear a variety of answers. Some willsay it is a regular server with added features, a

few describe it as a stripped-down box dedicated to aspecialized function, and still others believe the termrefers only to a network attached storage (NAS) box.

Not Your Average ServerThe typical server is configured to perform multiple func-tions. It operates as a file, print, application database,Web, or miscellaneous server. As such, it must have fastchips, more RAM, and plenty ofinternal disk space to cope withwhatever end users decide to dowith it.

Not so with a storage server. It isdesigned for a specific purpose, andthus configured differently. It maycome with a little extra storage or agreat deal.

"A general-purpose server typicallyhas five or less disks inside," saysGraham Lovell, senior director x64servers at Sun Microsystems. "Astorage server, on the other hand,has at least six, and more, usually 12to 24 disks."

Storage servers are normally individual units.Sometimes they are built into a 4U rackmount.Alternatively, they can consist of two boxes - a storageunit and a server located nearby. Both boxes can thenbe placed side-by-side in a rack. The Sun StorEdge3120 storage unit and SunFire X4100 server, for exam-ple, can be combined into a storage server and placedin a rack.

Apart from extra disks, what else is different about stor-age servers? In many cases, they come with a host ofspecialized services. This can include storage manage-ment software, extra hardware for higher resilience, a

range of RAID configurations and extra network con-nections to enable more users to be desktops to beconnected to it.

Just a NAS Box? Interestingly, some vendors define storage serverspurely in terms of NAS. A NAS appliance (also knownas a NAS filer) generally has a slimmed-down OS andfile system, and only processes I/O requests by themain file-sharing protocols. The big advantage of theNAS architecture is that it enables storage to be rap-

idly added by plugging the appli-ance into a network hub orswitch.

"As far as HP is concerned, a stor-age server is NAS," says JimHankins, product marketing man-ager for HP's NAS division. "Inessence, it is a dedicated file andprint server."

HP has a number of its ProLiantmodels available as general-pur-pose servers or storageservers/NAS filer - each has thesame basic hardware configura-tion. If licensed as a storage serv-er, the user may not run general-purpose applications on that serv-

er. If the same ProLiant server is being used as a regularserver, however, applications can be run on it.

In addition, HP's NAS-based storage servers have extrafunctionality built into the operating system - storage-specific management tools, "quota-ing" features, stor-age reporting capabilities, and a Web-based user inter-face that makes it easier to configure file and print.

So is NAS really just a storage server? The answervaries, depending on whom you ask. But it appearsthere is very little difference between them. NAS, itturns out, isn't really storage networking. Actual net-work-attached storage would be storage attached to a


Storage Basics: A Guide to the World of Storage Technology[ ]What Makes a Storage Server a Storage Server?

By Drew Robb

Jupiterimages

storage-area network (SAN). NAS, on the other hand, isjust a specialized server attached to a local-area net-work. All it does is make its files available to users andapplications connected to that NAS box - much thesame as a storage server.

"NAS is a marketing term," says Dan Tanner, an analystat storage consulting firm ProgresSmart. "NAS is reallynothing more than a file server, but specialized oradapted to the single purpose of serving files."

And what a marketing campaign it has been. Fromnowhere in the mid-1990s, Gartner projects the NASmarket will exceed $2 billion by 2008, with an annualgrowth rate of 9 percent. And those numbers don't takeinto account a new NAS flavor called the NAS gateway.These gateways act as a file-serving portal into a SAN:There are disk arrays in a Fibre Channel SAN that have astorage server on the perimeter acting as a NAS gate-way. This is a one way to marry up NAS and SAN assets.

"There are two flavors of storage servers," saysHankins, "NAS appliances that have the disk storage inthe appliance, and NAS gateways."

What’s Missing?While some vendors use the same box as a plain vanillaserver, others use a scaled-down version that is ade-quate for file serving. Steve Duplessie, senior analyst atEnterprise Strategy Group, defines a storage server asan optimized appliance designed to feed information,via a network, to a user or an application. As such, it isnot typically compute heavy, but it has been designedfrom the ground up to provide specific I/O capabilitiesalong with data protection capabilities.

A regular server has to be generic, it doesn't knowwhat kind of load demands it will have - gaming ismuch different than running a database, for example. Astorage server, such as a NAS box, is a contained appli-ance that does one thing really well, like file serving.

What does a "regular" server have that a storage serv-er doesn't? According to Duplessie, it typically hasmore processing power, more RAM, and a more gener-ic I/O structure and file system. As a result, most stor-age servers perform at 50 percent of the performanceof a regular server for the same function, he says. This trend toward specialized computing elements is farfrom new. TCP/IP routing, for example, was a function



The world of storage can be forbidding toa novice. Even veteran IT personnelmay be put off by the sheer volume of

new terminology and alphabet soup that hasevolved. Let's sample some basic terms:

Direct Attached Storage (DAS): The serverstores data on disks that are in the same box.Redundant Array of Independent Disks (RAID)is used heavily in this approach.

Storage Area Network (SAN): A collection ofcomputers and devices are connected over ahigh-speed network and are dedicated to thetask of storing and protecting data. Instead ofstoring data locally, each server sends dataacross the network to a shared pool of storage.

Disk Array: A large array of disks in one box, itis often used as part of a SAN to store data formultiple servers. These servers typically con-nect to the disk array using Fibre Channel.

Fibre Channel (FC): Optical fiber cables trans-mit data at high speed in a SAN. Fibre Channelis the transport protocol used for this purpose.

Network-Attached Storage (NAS): NAS sepa-rates data from applications by storing data onfilers attached to the LAN. Filers can sharefiles across multiple applications, platforms,and operating systems.

Internet Small Computer Systems Interface(iSCSI): This standard enables storage andretrieval at high speed (1 GB/second or higher)over regular IP networks.

- Drew Robb, Enterprise Storage Forum

StorageDefinitionsby Drew Robb

that every operating system ran - until Cisco came outwith a dedicated box that did it far better than hostingit on a general-purpose server.

"Any time you can optimize a function, it will be better[on a specialized box] than if executed on general-pur-pose gear," says Duplessie.

Dan Tanner, an analyst with the storage consulting firmProgresSmart, agrees with Duplessie's view that a stor-age server is a specialized server or appliance.

"The server OS is cut down to address purely printserver or file server functions, and often contains spe-cially tuned or enhanced code," says Tanner. "BeforeNAS came along, though, Microsoft said you could usea regular server for file serving."

But using a vanilla server for file serving could lead toproblems. Administering a general-purpose server ismore complex. Further, someone might be tempted touse the server for multiple functions. Dedicated storageservers, therefore, have become the norm.

Not surprisingly, Microsoft introduced Windows StorageServer 2003 to distinguish it from general servers run-ning the Windows 200x operating system. WindowsStorage Server 2003 is a dedicated file and print serverbased on Windows Server 2003 and tailored to net-worked storage. It supports file serving and backup andreplication of stored data. It can also be used to consoli-date multiple file servers into a single box.

Storage Servers vs. Disk ArraysJust as there is some confusion between ordinaryservers and storage servers, there is also sometimes amisunderstanding between storage servers and disk

arrays. Exactly where does one end and the otherbegin? A storage server can have as many as 24 disks -enough to qualify as an array. Disk arrays, however, canhave hundreds of disks. So where do you draw theline?

"A storage server is usually standalone and not con-nected to other servers," says Lovell. "Multiple servers,however, typically connect to a disk array."

Disk arrays, too, often connect to a server that could bestyled a storage server. The storage server is the intelli-gence that goes in front of the array. In this arrange-ment, the server can manage several tiers of storage. Itcan even arrange the replication of data from one tierto another.

"A storage server serves the storage, and the disk arrayis the storage," says Tanner. "Using a storage serverlets you use multiple or different arrays."

Duplessie further separates the two terms.

"A storage server typically speaks to files and talks topeople or applications over Ethernet," says Duplessie."A disk array is a low-level block device that onlyspeaks to an operating system." ■



Storage Server Differentiators

• Lots of disks (12-24)• A standalone unit• Preinstalled software apps to manage

the data or storage-specific peripherals• Usually less powerful than its

pre-installed counterparts

30 percent of 288 storage professionals surveyed said their companies' security policiesdid not include storage systems. -- Enterprise Strategy Group

“”

Storage is an immense and complex universe.Once you enter, your mind is soon swimming instrange, even alien concepts. Therefore, it is best

to stick to what you know and keep it very simple -especially at the start.

One obvious way to avoid complexity is to use the serv-ices of a storage service provider. These are firms thatlease storage from their own data centers and otherservices. Colorado Software Architects, for example,offers 1Disk.com. Sun, Arsenal Digital, and Iron Mountainare among the companies with similar services.

The advantage of a storage provider is that the ven-dor provides a variety of storage options for a fixedcost. This is a handy way to addstorage capacity or meet regu-latory compliance/archivingrequirements without having tobuild new infrastructure.

Of course, simplicity can betaken to extremes (i.e., attempt-ing to pass the entire storageburden to an external source orkeeping everything stored onthe same old servers using big-ger and better disks). Such astrategy eventually runs into awall; there is so much datastored on so many servers that itbecomes impossible to manage.

Beyond DAS, then, whereshould the rookie storage guygo to ease his woes? Initially, at least, it might besmart to start with NAS and avoid SANs. At its core, aNAS filer is simply a specialized type of server thatconnects to the network. Storage is rapidly added byplugging the appliance into a network hub or switch.The likelihood is that the server administrator will runinto very little that is new to him by buying a NASbox. Lower-end models that are relatively easy to use

are available from Network Appliance, SnapAppliance (now owned by Adaptec), and HP.

The drawback of NAS is that filers and servers share thesame LAN. As a result, network performance may even-tually be affected. When that juncture is reached, itmay be remedied by upgrading the LAN and addinghigher-grade NAS equipment. A more long-term solu-tion would be to roll out the first SAN.

Simple SANman SaysUndoubtedly, the land of the SAN can be forbidding.Continuing with our theme of simplicity, the transitionto a SAN can be made smoother by beginning with

rapidly maturing iSCSI technol-ogy. iSCSI allows the establish-ment of a SAN over an IP net-work. Thus, the IT departmentdoes not need to learn newprotocols or add new skill setsto create a SAN. This also hasthe advantage of being muchless-expensive than an FC SAN.

Super-Size ItiSCSI is especially appropriatefor companies with IP back-bones capable of handlinggigabit traffic. While the tech-nology is improving rapidly, itdoesn't offer the same speedor capacities as a heavy-dutyFC SAN. Similarly, SANs offerhigher speeds and throughput

than NAS systems. To do this, they offload data trafficto a separate network for storage devices.

On the negative side of the ledger, however, SANs mayhave difficulty supporting multiple operating systemsand platforms. In addition, some users complain aboutbeing unable to integrate SAN solutions from differentvendors.


Storage Basics: A Guide to the World of Storage Technology[ ]Storage Strategies Made Simple

By Drew Robb

Jupiterimages

Choose WiselyThe basic strategy for storage is to try to stick with thefamiliar. NAS and iSCSI are good starting points forcompetent IT departments already familiar with IP net-working. FC SANs, on the other hand, should probablybe avoided unless you have very large capacity andrequire the highest possible performance.

If so, it is best to recruit a dedicated storage team towrestle this beast and bend it to your corporate will.

Although the cost and complexity are greater in theshort term, the potential long-range payoff is greaterthan with NAS or iSCSI.

And for those that just don't want to involve them-selves in yet another IT skill set, managed storage serv-ices now cover the entire spectrum. Sometimes it is justless-expensive, easier, or faster to call in the profession-als and leave everything to them. ■



Given the emphasis administrators and corporatemanagers place on IT security, it's hard to imag-ine an environment in which security implemen-

tations are not a primary concern. As such, many oftoday's network IT administrators carefully consider allaspects of security when deploying and managing theirnetworks.

Despite all the well-documented threats and mediaattention, however, there is no shortage of networks thatare still operating with minimal and poorly implementedsecurity measures. This can be due to lack of knowledgeabout the real risks to data security,unaddressed vulnerabilities, andsometimes to a false sense of securi-ty due to reliance on inadequatesecurity strategies.

Storage networking technology hasenjoyed strong growth in recentyears, but security concerns andthreats facing networked data havegrown equally fast. Today, there aremany potential threats that are tar-geted at storage networks, includ-ing data modification, destructionand theft, DoS attacks, malware,hardware theft and unauthorizedaccess, among others. In order for aSAN to be secure, each of thesethreats must be individuallyaddressed. Fortunately, many of thesecurity practices and protocols used to address tradi-tional network vulnerabilities also help ensure the avail-ability of storage networks by reducing common securi-ty threats.

At the ground floor of any security strategy are somebasic security concepts, including authentication, authori-zation, encryption (confidentiality), integrity, accountabili-ty and access control. We'll start with access control.

Access Control Access control is a cornerstone concept when design-

ing a secure network environment. Access control is allabout controlling who can and cannot access a net-work, a resource, a folder or file.

In order to effectively secure such resources, you mustcarefully consider and control the level of access grant-ed to each network user and then deploy strategies toensure that only required users actually have resourceaccess. It is a fundamental concept, and the foundationfor a strong and secure network environment.

There are several types of access control strategies,including mandatory access control (MAC), discre-

tionary access control (DAC), androle-based access control (RBAC).

MAC represents the tightest formof access control. In this strategy,security policies prevent the cre-ator of any information from con-trolling who can access or modifytheir data. Instead, administratorsor managers maintain control overwho can access and modify data,systems and resources. Mandatoryaccess control systems are com-monly used in highly secure net-work environments such as militaryinstallation or financial or medicalinstitutions.

MAC secures information andresources by assigning sensitivity labels on objects andcomparing this to the level of sensitivity a user isassigned. This label is a kind of confidentiality stamp;when a label is placed on a file it describes the level ofsecurity required to access that specific file and willonly permit access by files, users and resources with asimilar or lesser security label.

MAC assigns a security level to all information, andplaces security clearance to each network user toensure that all users only have access to that data forwhich they have security clearance. For example, usersmay be assigned a security label such as top secret or


Storage Basics: A Guide to the World of Storage Technology[ ]Storage Security Basics

By Drew Robb

Jupiterimages

confidential, and data and resources are classifiedaccordingly. MAC restricts access to objects based on acomparable sensitivity between the user-assigned lev-els and the object-assigned levels.

The administrator or the operating system policy doesnot force discretionary Access Control (DAC); instead,an object's owner controls access. In a DAC model, if auser creates a folder, that user decides who will haveaccess to that folder.

DAC is associated with an access control list (ACL). TheACL maintains information on the rights a user has to aparticular system object, such as a file, directory or net-work resource. Each object has a security attribute thatidentifies its access control list and the list has an entryfor each system user with associated access privileges.The most common privileges include the ability to reada file (or all the files in a directory), to write to the file orfiles, and to execute the file (if it is an executable file orprogram).

Microsoft Windows 2000/2003/XP, Linux, UNIX and MACOS X are among the operating systems that use accesscontrol lists, although the list is implemented differentlyby each operating system. In Windows NT/2000/2003,an ACL is associated with each system object. Each ACLhas one or more access control entries (ACEs) consistingof the name of a user or group of users. The user canalso be a role name, such as "secretary" or "research."For each of these users, groups, or roles, the access priv-ileges are stated in a string of bits called an access mask.The system administrator or the object owner typicallycreates the access control list for an object.

In a role-based access control (RBAC) configuration,access decisions are determined by the roles that indi-vidual users have as part of an organization. In anyorganization network users are assigned specific rolessuch as marketers, salespeople, managers, secretariesand so on. Users with similar roles are grouped togeth-er, and access control is determined by the role those



Back in the days when storage meantdirect attached storage (DAS), storagesecurity was included in overall IT

security. But as storage architectures havedeveloped with the introduction of high-speed,high-capacity Fibre Channel-based storagearea networks (SANs) as well as more tradition-al Ethernet-based network attached storage(NAS) systems, storage security has become adiscipline in itself. Neglect it at your peril.

The starting point for a systematic approach tostorage security, according to Sal Capizzi, asenior analyst at Boston, Mass.-based YankeeGroup, is to take stock of the various types ofdata being stored and classifying it accordingto how important it is and how costly it wouldbe to the business if it were lost or stolen. Thenfor each classification, appropriate securitypolicies should be set.

The next step, Capizzi says, is to enforce pass-word and World Wide name identification (forFibre Channel) and logical unit number (LUN)authorization to ensure that only authorizedusers, devices or applications can access data,and to implement LUN masking so that partic-ular storage volumes can only be seen byauthorized users, devices or applications.

Ensure that all actions, accesses and changesto data are logged to provide a clear audit trailof who did what to which data from where, andwhen. Without such logs it is very hard to tell ifor how data has been compromised.

Finally, don't neglect the boring obvious stuff:Use anti-virus, and anti-spyware software and asuitable firewall, disable unused ports, changepasswords frequently, and so on.

-- Paul Rubens, Enterprise Storage Forum

StorageSecurityby Paul Rubens

Twenty percent of companies do not know or are not in a position totell if their storage security has been

breached.

“

”

users have on the network. Role-based access requiresa thorough understanding of how a particular organiza-tion operates, the number of users and their exact func-tion in that organization.

Access rights are grouped by role name, and the use ofresources is restricted to individuals authorized toassume the associated role. For example, within aschool system, the role of teacher can include access tocertain data, including test banks, research material,memos and related material. School administrators mayhave access to employee records, financial data, plan-ning projects and more.

When a user is associated with a role, the user shouldbe assigned only those privileges necessary to do theirjob. This is a general security principal known as the"least privilege" concept and applies to all access con-trol methods. In a role-based scenario, when someoneis hired for an organization, their role is clearly defined:teacher, secretary, sales, marketing, manager, etc. Anew account is created for the user and then placed ina group with those with the same role within the organ-ization. Individual permissions do not need to be set;rather, the level of access control is inherited from thegroup in which they are placed. As an example, if anew teacher is hired for a school, the user account isplaced in the Teachers Group. Once in the group, thenew employee will inherit the same level of access asthose already in the Teachers Group.

Role-based access control is actually a form of MAC,since access is dictated by an administrator and the cri-teria for object access in not in the hands of the owner.

Authentication, Authorization andAccountability Poor user authentication and authorization are one ofthe most common weaknesses in networks, and stor-age area networks are no different.

Poor user authentication and authorization are impor-tant concepts in network security. Authentication refersto the process by which you verify that someone is whohe or she claim they are. This traditionally involves ausername and a password, but can include any othermethod of demonstrating identity, such as a smart card,biometrics, voice recognition, fingerprints, and so on.Authentication is a significant consideration for networkand system security and an important part of maintain-

ing secure access control. Authentication security iscontrolled through policies and protocols. In an IPLAN/WAN environment, CHAP, EAP and MS-CHAP areexamples of authentication protocols. There are alsoauthentication protocols unique to a SAN environment,including both a secret key design with DH-CHAPauthentication and public authentication with FCAP(Fibre Channel Authentication Protocol).

Authorization refers to the process of determining if auser, once identified and authenticated, is allowed tohave access to a particular resource. This is usually deter-mined by finding out if that person is a part of a particu-lar group that provides the correct permissions, rights orrequired level of security clearance to access a resource.

Accountability refers to the tracking mechanisms used tokeep a record of events on a system. One tool oftenused for this purpose is known as auditing. Auditing isthe process of monitoring occurrences and keeping alog of what has occurred on a system. It is largely up tothe administrator what types of events should be trackedand which should not. By tracking events on a system, itis hoped that attempts to access the network or other-wise compromise data will be recorded and prevented.

Confidentiality and Integrity In any security strategy, protocols are needed to pre-vent data from being read by intruders (confidentiality)and others to determine if data has been tamperedwith during transit (integrity).

To prevent data from being read, encryption is used.Encryption takes raw data and scrambles it in such a waythat it is unreadable without the key. If the correct key isnot available, the stolen data maintains its confidentiality.As an example, within IPSec, the Encapsulating SecurityPayload (ESP) protocol can encrypt data sent over FibreChannel links. Regular Ethernet communications can alsouse IPSec encryption or other protocols such as theSecure Sockets Layer (SSL) protocol. All encryption pro-tocols are designed to make intercepted data unread-able to ensure confidentiality.

Integrity refers to the checking of data to ensure thatdata has not been tampered with or modified in anyway. As an example, during the IPSec key exchangeprocess, initial negotiations use one of two integrityverification methods, the message digest 5 (MD5) orSecure Hash Algorithm (SHA), to ensure that data hasnot been tampered with during the process. ■



With the price per gigabyte of storage comingdown rapidly, that line item is no longer theoverriding consideration for most storage

budgets. While that is some relief for storage users, inother ways it creates a new problem: how long shouldyou wait for storage to get faster and cheaper beforeyou buy?

Add to that the complexity of upgrading to new tech-nologies - 2Gbps vs. 4Gbps Fibre Channel, for exam-ple, or SAS vs. SATA, SCSI or Fibre Channel - andyou're confronted with an array of planning and budg-eting issues when it comes time to upgrade or replaceyour storage architecture.

Budgeting for storage is notjust about buying more den-sity or the latest cool stuff; itis about determining yourneeds based on availabletechnology, and making surethose requirements are met.

The important issues to con-sider when budgeting forstorage are:

1. How will a new technol-ogy integrate into the cur-rent environment? 2. Will this technology meet user requirements forperformance and reliability? 3. How does this new technology affect O&M (opera-tion and maintenance) costs?

Integration Integration of technology into the current environmentis a large problem for several reasons. Let's take a real-world example from an actual site. They have serversfrom one vendor and storage from another. The stor-age vendor can provide a new storage infrastructurethat will support 4Gb Fibre Channel RAID controllers,4Gb Fibre Channel switches, and other storage compo-nents. That all sounds great, but can the the server side

support the 4Gb architecture?

This is a big question that should be asked of everyhardware vendor. A standard PCI bus running at fullrate supports 536 MB/sec, but many PCI buses do notsupport this full rate, and even though the situation isbetter, the same is also true for a PCI-X bus running atapproximately 1.1 GB/sec (twice the PCI rate). A two-port 2 Gb HBA can require up to 800 MB/sec (200MB/sec for each port reading and 200 MB/sec for eachport writing). Therefore, a standard PCI bus cannot sup-port two-port HBAs running at 2 Gb, which would bethe same as one port at 4 Gb.

From a failover point of view, having two ports with 2Gb provides greater redun-dancy if an HBA port fails,which is more commonthan both ports failing. Thisassumes that you have anHBA failure and not a PCIbus failure. In the case ofPCI-X, a two-port 4 GbHBA far exceeds the PCI-Xbus bandwidth, (1.1 GB/secfor PCI-X, and two ports ofa 4 Gb HBA require 1.6GB/sec for full rate), so per-formance is far closer to

that of two ports of a 2 Gb HBA.

All of these performance numbers assume that the I/Obeing done is streaming I/O. If it isn't, then why evenconsider 4 Gb HBAs and infrastructure in the firstplace? Yes, you can get improved IOPS performancewith 4 Gb HBAs from a larger command queue, but theperformance improvement is not that great and is oftenvery workload-dependent. Ranges I have seen are from0%-20%, but your mileage may very. This improvedperformance is surely not a justification to run out andbuy a 4 Gb infrastructure.

The bottom line is that any site considering 4 Gb tech-nology must make sure that the servers can supportthis new performance level. More often than not, large


Storage Basics: A Guide to the World of Storage Technology[ ]Storage Budgeting Tips

By Henry Newman

Jupiterimages

servers lag in bus technology, given the large lead timeit takes to design the complex memory interconnectsto the bus and the availability of new bus technology.You can buy PCI-Express bus technology from Dell onone, two and four CPU systems, but try to find that onlarge (greater than 16) multi-CPU servers today.

User Requirements User requirements should be a major driver of technol-ogy upgrades. Many organizations do not have a goodhandle on what the user application profiles look like,what the growth requirements are, and worst of all,whether the system is configured and tuned for thoseapplication profiles. This lack of understanding of theenvironment can lead to poor decisions on what hard-ware and software is needed.

One system I recently reviewed did not have an emula-tion or characterization of their workload. This is espe-

cially important for large sites. Without this information,how could this large site test patches for performancedegradation (yes, it happens all too often), test newtechnology to measure performance improvements, ortest increases in workloads to see if the system canhandle them?

User applications and requirements should be a largecomponent in any decision to upgrade technology. Ifyou do not know what users are doing with the system,how do you know what they need today, let alone planfor the future? This situation often turns into a fire drillwhen the system is overloaded, and management startsthrowing money at the problem instead of executing amaster plan for technology infrastructure upgrades.

O&M Considerations Technology maintenance costs almost always follow thesame pattern:

• The cost of O&M for new technology is high forearly adopters.

• Over the next 6 to 18 months, the cost drops asthe technology is more widely adopted. • The cost continues to drop, and drops sharplywhen a technology replacement is released, until... • The cost skyrockets as the vendor tries to phase outthe technology. This value is far greater than the orig-inal cost of maintenance, and sometimes I have seenit go as high as five times greater, since the vendorno longer wants to support the technology becauseof its cost and wants you to upgrade.

This is the general lifecycle for O&M costs. It makessense given vendor costs, and unless technology trendschange, the pattern is likely to continue.

One other area that should be considered is the per-sonnel cost to the organization of supporting old hard-ware and software. You're not likely to find a new hirewho knows how to work on Fibre Channel arbitratedloop HBAs, RAIDs and switches, and finding training

course for that hardware isn't an easy task either. Justrecall the frantic search for mainframe COBOL pro-grammers for Y2K - a clear example of personnel oper-ations costs becoming unreasonable.

Conclusions The issues addressed here are the ones that drive thehigh cost of storage changes. Most sites know whattheir physical storage growth will be, or at least whatthe budget will allow them for physical storage growth.The major cost items are not adding a few trays ofdisks with 146 GB drives or swapping out 36 GB drivesfor 300 GB drives; the major cost drivers are the infra-structure. The real question is how do you determinewhat you need, how much it is going to cost, and howto fit it into your current environment.

One pitfall: sites think they can just jump into new tech-nology without fully understanding the whole data path(the path from the application to the operating systemto the HBA/NIC to the storage devices). Plugging 4 Gb



Fewer than 25 percent of either Unix-/Linux- or Windows-based IT organizations hadtheir own storage management team at the end of 2004. By the end of 2006, however,

that number is expected to soar above 75 percent.

“

”

HBAs in current servers into a 2 Gb storage infrastruc-ture does not generally improve performance unlessyou are aggregating the performance of multiple RAIDcontrollers and multiple hosts. The science (some callthis an art, but it is really based on scientific analysisand study of the data path) of determining what usersneed and when they will need it is the process ofbudgeting for storage.

You need a full understanding of: • Your current environment, including the perform-ance level that environment can support today andthe performance level that environment can supportgiven technology trends; • User requirements for performance and growth,including the current workload and the trend line forgrowth (performance mapped to expected new tech-nology); and • Your current and future O&M costs. Don't wait untilyour maintenance contract ends to find out that the

cost has sky rocketed - technology maintenance costsfollow a pattern.

Budgeting for storage is considered by many to be acomplex problem, but it's not very complex if the linesof communication between the affected groups areopen and free-flowing. The key is to have the data -seeing the future does not require a crystal ball, justsome understanding of what you have and what youuse, mixed in with a bit of history. ■

About this informationThis content was adapted from EarthWeb's EnterpriseStorage Forum Web site. Contributors: Dan Muse, Paul Shread, Drew Robb, Mike Harwood, and Henry Newman.

Copyright 2007 Jupitermedia.



JupiterWeb eBooks bring together the best in technical information, ideas and coverage of important IT trendsthat help technology professionals build their knowledge and shape the future of their IT organizations. Formore information and resources on storage, visit any of our category-leading sites:

www.Enteprisestorageforum.comwww.internetnews.com/storagewww.linuxtoday.com/storagewww.databasejournal.comhttp://news.earthweb.com/storagehttp://www.internet.com/storage

For the latest live and on-demand Webcasts on storage, visit: www.internet.com/storage

Documents

Storage Basics