5 ways to whip your storage into shape
Swelling data volumes and more online operations demand a new level of fitness for government storage systems
- By John Moore
- Aug 21, 2005
Rare is the government data storage operation that wouldn't benefit from a little improvement-minded attention. Agencies typically have ample opportunity to boost efficiency, cut costs and better protect data. The project vignettes below illustrate storage management improvement plans that can help organizations achieve those worthwhile objectives.
Too much data overwhelming traditional backup routines
Fitness plan: Virtual tape
The virtual tape approach, in which disks emulating a tape drive are used for backup instead of actual tape, attracts organizations that face backup windows pinched by growing data volumes and a need for quicker data recovery time than tape can provide.
For years, the higher cost of disk storage has restricted it to serving as the primary online platform, combined with less expensive but slower tape handling backup and archival tasks. But because the price difference between disk and tape has narrowed, agencies can now incorporate disk-based storage into their backup and recovery strategies. Several vendors provide virtual tape products that use inexpensive Serial Advanced Technology Attachment disks. The disk solutions' main advantage over tape is speed.
The Brookhaven National Laboratory is one agency that has adopted virtual tape. Last year, it deployed a disk-based virtual tape library from Sepaton.
Andrew Ferguson, manager of enterprise operations at Brookhaven, cited the speed of backup and restoration as the main reasons for pursuing virtual tape. The lab now conducts backups in less than seven hours instead of the 12 it took using conventional tape.
In addition, Ferguson said, "restores are very fast, especially if you need to mount more than one tape." The tape-mounting process can add nearly 15 minutes to restoration, he said.
"This has increased backup speed tremendously on the large stuff we back up," Ferguson said. "We haven't had to do a large restore, but the smaller restores have been successful."
Rich Dickson, vice president of channel sales at Sepaton, said the company has encountered a number of government agencies deploying virtual tape or considering an installation. "They...can't keep buying more and more tape devices to try to shrink the windows," he said.
Brookhaven uses a Sepaton virtual tape library, which provides 24.5 terabytes of raw storage. Ferguson said the laboratory integrated the product into its Dell storage-area network (SAN) without problems. Brookhaven's backup software, EMC Legato's NetWorker, picked up the Sepaton devices as it would a conventional tape library.
"It looked like another library," he said. "Setting up was relatively easy."
High cost, special skills associated with Fibre Channel SANs
Fitness plan: IP storage networks
IP-based SANs have taken awhile to catch on, but adoption has grown within the past year among government agencies.
The attraction: The storage technology lets organizations deliver block-level storage via nearly ubiquitous Ethernet networks based on TCP/IP. IP SANs save organizations money because most IT shops already possess the necessary infrastructure and personnel. Fibre Channel SANs, in contrast, require a separate network and technicians who specialize in the technology.
The Denton Central Appraisal District in Denton County, Texas, operates a Fibre Channel SAN and an IP SAN. The Fibre Channel SAN, from Dell and EMC, handles the organization's most critical production databases and supports highly transactional applications, said Brad Green, the district's director of information services.
But the IP SAN, from StoneFly Networks, provides a cost-effective means for accommodating the district's overall data growth, which has soared as high as 300 percent per year.
The IP SAN, the installation of which began in late 2003, also addresses the district's need to deal with periodic business surges that require extra storage capacity. The organization issues annual appraisal notices to property owners who have three months to protest before tax assessments become final.
"Our business really ramps up after we send out the appraisal notices," Green said.
In one case, the district needed an extra database server. StoneFly, he said, was able to provision disk storage for that platform in a matter of minutes without his organization "having to buy a Fibre [Channel Host Bus Adapter] or bring down a box and install an HBA."
The district uses the IP SAN to run all backups and plans to use StoneFly's Replicator software to asynchronously copy data to a disaster recovery site.
The IP SAN is easy to use, decreasing training and staffing expenses. "It's a lot easier to use than a Fibre SAN," Green said.
"Governments are one of the more natural places for this technology," said Bob Peyser, vice president of marketing at StoneFly, noting that IP SAN products have found traction among value-oriented customers. "Moving to an IP SAN is basically a way to lower the amount of effort it takes to provision storage."
Sprawling data stores tax equipment and staff
Fitness plan: Storage consolidation
Many organizations have attached storage to myriad servers
to keep up with the rapid growth in data. Such environments become difficult and expensive to manage. Consolidation offers an alternative approach for government agencies seeking scalability.
Pierce County, Wash., which includes Tacoma and Mount Rainier, embarked on a consolidation project last year. The county maintains a mountain of geographic information systems (GIS) data and has a potential user base that includes 35 agencies and 744,000 residents.
The declining cost of orthographic imagery had spurred demand for the county's GIS holdings, said Linda Gerull, Pierce County's GIS manager. "The files are huge and take up an enormous amount of space," she said. "Managing all that and storing all that is a huge challenge."
The county had been storing its data on direct-attached storage (DAS) devices and an underutilized SAN. County officials determined that adding new servers, storage devices and the personnel to manage them was an untenable solution.
Instead, county officials set about the task of moving their data from decentralized, heterogeneous storage devices to a centralized SAN from IBM. Components include IBM DS4500 disk arrays and Tivoli Storage Manager software, said Rick Schieche, the Pierce County project's technical co-lead. Tivoli Storage Manager coordinates backups and disaster recovery.
Tricia Jiang, technical attaché at IBM Tivoli Storage Systems, said consolidation makes storage easier to administer and gives customers better availability.
The county's central data storage pool offers users ready access to data and eases the administrative burden on the county's limited IT staff. In the past, a full-time staffer "did nothing but move data files around" because storage was tied to individual servers, Gerull said.
The new storage system also lets the county keep its data online, which speeds file recovery. Previously, "our support staff was spending a lot of time trying to pull data files off of tape," she said. IT personnel can now recover files in 10 minutes as opposed to one hour or longer.
The ability to manage the county's expanding storage needs without increasing staff saves money. Gerull said the county will save $3 million over four years.
The central SAN also will help the county handle its rapidly accumulating data holdings. The county bought the IBM SAN, which has 3 terabytes of storage, last year. But with the recent addition of Light Detection and Ranging tools to measure terrain data and updated orthographic imagery, county officials "realized we needed another 3 terabytes that will last us through 2006," Gerull said. The new SAN environment offers the additional headroom.
Inefficient use of available storage space and burdensome administration
Fitness plan: Storage virtualization
The practice of virtualization centers on the creation of a single pool of logical storage, which can be provisioned as needed. Benefits include improved use of storage devices and easier management.
Tricare, the military's health care provider, implemented storage virtualization through consolidation. The organization combined regional operations. In the western United States, five Tricare regions became one.
What had been known as Tricare Region 9, based in San Diego, became the hub for the Western Region. The Region 9 data center was relocated from Balboa Naval Hospital to a larger facility in downtown San Diego. Data from Region 9 and the other regions was migrated to a Xiotech Magnitude 3D SAN. Region 9 was using an older Xiotech SAN, while the other regions used DAS.
Tricare contractor IntelliDyne managed the migration project. Chris Morgan, a systems engineer at IntelliDyne at the time, said the new Xiotech SAN lets storage administrators carve out space for use with a particular server.
Maj. Kevin Seeley, Western Region chief information officer, said "the ability to provision virtual disk space provided the extra flexibility and scalability which enabled us to rapidly stand up our new data center operations."
Customers find virtualization attractive because they can "provision on the fly and reprovision as they need to," said Rob Peglar, vice president of technical solutions at Xiotech.
Xiotech calls each allocated space a VDisk, or virtual disk. Morgan said a server sees a virtual disk as attached storage.
At Tricare, the virtualization feature smoothed the migration process. The new data center was to be set up in three days. To speed the task, Morgan and his colleagues configured disk space for the data center's servers before the move to the new downtown San Diego office space. The task was completed using one server to create templates for different servers Web, e-mail and database servers, for example.
"We just used one server and pointed it at each disk space," Morgan said. That approach eliminated the need to unpack multiple servers to prepare them for the move. When the servers arrived at their destination, the Magnitude 3D had server images ready on virtual disks.
"That saved a lot of time," Morgan said.
Inadequate resiliency exposes data and operations to unreasonable risk
Fitness plan: Business continuity through system redundancy
Organizations aim for continuous data availability, but reaching that destination may call for a stepwise approach.
The Air Force's 45th Space Wing has been building a disaster recovery and backup capability for the past three years. Business continuity is an important issue for this group. The wing's two nodes Patrick Air Force Base and Cape Canaveral Air Force Station are located in Florida's hurricane alley.
The initial step was setting up clustered servers for service continuity and a dual-fabric SAN. In 2003, the wing operated a SAN at each location, but it has since consolidated them into one large SAN. Data is copied between the base and Cape Canaveral.
The Air Force SAN environment includes Brocade's SilkWorm 3800 switches soon to be upgraded to the 4100 platform Computer Associates International's BrightStor SAN Manager and Enterprise Backup, and EMC's Clariion storage arrays.
Another phase of the wing's disaster recovery and backup plan started about a year ago. The Air Force unit uses virtual tape libraries from Network Appliance. The virtual tape solution provides throughput of 2.5G to 5G per minute, said Glenn Exline, manager of advanced technology at Computer Sciences Raytheon, which manages IT for the 45th Space Wing.
In addition, the virtual tape "eliminates a lot of the problems with physical tape libraries," Exline said. He cited robotic problems and dirty tape heads. The organization still uses physical tape libraries from Qualstar, however.
The 45th Space Wing's next data-protection move came in November 2004. The organization adopted a Microsoft Exchange cluster environment. A cluster exists at each of the wing's locations. The Air Force unit uses EMC's SnapView and SAN Copy to replicate data between the two sites.
Exline advised organizations to take on SAN and business continuity projects in manageable chunks. "If you do it the other way, where you try to implement the SAN and Exchange information and data replication in one fell swoop, typically you'll find that you've bit off more than you can chew," he said.