Store more with less

Agencies face similar storage challenges,so they often ask systemsintegrators how they can storemore data in a smaller space while keepingit easily accessible.Disk arrays are oneanswer. At its mostbasic, a storage diskarray is a collection ofdisk drives withcache memory, controllers,power suppliesand other supporting hardware.The current generation of arrays alsooffers a smorgasbord of speed, capacity,data protection, security, replication andspace-saving tricks. And many productsoffer savings in power and cooling ?important considerations given the costof energy and increasing mandates forgreen computing.At the same time, the space needed toaccommodate disk arrays is shrinking."You can now double, triple or quadrupleyour data capacity in the same footprintusing less energy," said Greg Schulz,founder and senior analyst at consultingfirm StorageIO Group.Finally, many government agencies aretrying to satisfy multiple requirements. Insome cases, retrieval speed is crucial. Inother situations, particularly when data isstored online, drive speed might be lessimportant because Internet connectionswill determine how quickly files can beaccessed. An agency customer mightwant to trade disk speed for smaller size,lower power, better cooling or lower cost.The agency should focus on what it istrying to achieve, said R.B. Hooks, chieftechnology officer of Sun MicrosystemsFederal's Storage Group.Fortunately, new developments in technicalprotocols are helping disk arraysprovide data more easily. A trend towardfunctional convergence is combiningFibre Channel and Internet SCSI drives,Common Internet File System andNetwork File System drives,and Fibre AttachedTechnology Adaptedand Serial AdvancedTechnology Attachmentdrives, said AndrewReichman, a senioranalyst at ForresterResearch.In other words, systems integratorshave numerous options for meeting theircustomers' objectives. Do they want toshrink their data centers? Then use high-capacitydrives. Do they need quickeraccess to data? If so, recommend thefastest products. Do they want to savemoney? Avoid the biggest and speediestmodels and choose the next-lowest tier ofcapable drives.Most disk arrays use some form ofRedundant Array of Independent Diskstechnology, which achieves higher reliabilitythan a single disk by distributingdata across two or more drives. Also, thetechnology allows the disks to send datafaster than individual drives.Consolidating resources is a primarygoal of many agencies and departments,partly to reduce costs and partly to simplifymanagement."Customers often want fewer data centersor just one," said Kyle Fitze, directorof marketing at Hewlett-Packard'sStorage Platforms Division. Given theprevalence of concentrated storage, consolidationhas become not only feasiblebut also fairly easy.Lean provisioning is one strategy fordoing more with less. Systems often allottoo much space to applications, which canwaste resources. Rather than reservingstorage space for an application or server,lean provisioning makes that space reservationconditional. For example, if twoapplications each supposedly require 10T,lean provisioning would grant each oneonly 6T, with perhaps another 4T inreserve. Those applications likely will becontent with their lean rations, and thesystem will have space for other needs.How likely is lean provisioning to helpyour customer? It depends on the applications,data and operating system."Typically, Windows has 25 percent to30 percent usage, Unix has 35 percent to45 percent usage, and mainframes about80 percent usage," Hooks said. That leavesa lot of wasted space.Vendors implement lean provisioningin various ways. For example, Sun offersthe StorageTek 9900 Software Suite forvirtualization and dynamic provisioningand the Solaris ZFS 128-bit file system,which doesn't consume storage until datais written.Data deduplication is another strategy.It reduces or eliminates redundant information.For example, if you send ane-mail message with an attachment to aco-worker, there are now two copies ofthat attachment. Rather than store bothcopies, deduplication tools store only oneand point to that one copy for all otherreferences."While common in backup systems,deduplication is now in primary systems,"Reichman said. The technology saves onthe space and bandwidth needed to movethat data around.In addition, data compression canreduce the raw size of stored files.Integrators should also consider tieredstorage, in which different tiers of disksare used for different purposes."You can set the rules for which databelongs on which tier," said Mark Peters,an analyst at Enterprise Strategy Group.Most agencies are required to maintainbackups of their files, preferably at anotherlocation. Many hard-drive arrays performthat replication automatically."The idea is disaster avoidance withoutdisruption, not disaster recovery," Schulzsaid.However, agencies often discover thattheir disaster recovery plans don't workwhen it's too late. For example, CharlotteCounty, Fla., officials had to physicallymove servers to a safe location duringHurricane Charley in 2004."Many of our servers were runningwithout air conditioning, which is notideal," said Mark Ramsey, manager ofthe county's information technologyoperations.Securing data is as important as backingit up. Accordingly, some disk arraysnow offer the option of encrypting all dataand digitally shredding it if a drive isremoved without authorization."This can help any agency with issues ofcompliance or chain of custody," saidDave Egan, senior vice president of storageat Fujitsu Computer Systems.DriveTrust technology ? a joint effort byAdvanced Micro Devices, HP, IBM, Intel,Microsoft, Seagate Technology and Sun? features firmware that encrypts drivesand enables secure erasure. In addition,manufacturers such as EMC and Fujitsuoffer encrypted drives.However, if an agency customer choosesdrive encryption, integrators shouldrecommend an investment in specializedsoftware that keeps track of encryptionkeys so the customer doesn't have to.To accommodate future needs forexpansion, consider vendors that offercompatible disk arrays at different levelsof storage.Fortunately, transmission protocolsare also changing. The hot one thesedays is Fibre Channel over Ethernet(FCOE). Many data centers already useEthernet for their TCP/IP networks andFibre Channel for their storage-area networks.By adopting FCOE, agencies canrun Fibre Channel traffic over theirEthernet connections without the needfor special Fibre Channel cabling or theless-well-known iSCSI protocol."This simplifies implementationbecause most staff [members] are alreadyfamiliar with Ethernet," Peters said.