The Fibre Channel (FC)-connected disk drives that populate midrange arrays may soon give way to a new generation...
of Serial-Attached SCSI (SAS) drives, says Tom Georgens, president and CEO at Engenio, whose midrange arrays are resold by vendors like IBM, SGI and StorageTek. Pending review of SAS drives' reliability, Engenio may ship a midrange SAS-based product as early as 2007. "It's my clear preference to do it sooner rather than later," says Georgens.
At issue is customer demand for high-capacity SATA drives used to create a second tier of low-cost capacity in existing midrange arrays. SATA drives can natively plug into a SAS backplane, but if you want to put a SATA drive in an FC-based enclosure, you need to use FC-to-SATA protocol conversion hardware. That adds approximately $400 to $500 to a disk shelf, says Bob Brencic, senior director of switch marketing at FC component maker Emulex.
Furthermore, SAS drives, which will likely replace parallel SCSI disk drives in servers, will ship in greater numbers than FC drives. Unless FC disk-drive vendors significantly reduce their prices, SAS drives promise to be much cheaper.
But to make the switch from FC to SAS enclosures, array manufacturers will have to ditch significant amounts of FC infrastructure, engineering and test equipment. "Why throw out something robust that you have a lot of knowledge about for something that's just full of promises?" asks Brencic.
Rather than let that happen, Emulex and FC players Agilent, Brocade, Hitachi Global Storage Technologies, McData and Seagate are working on adding native SATA connectivity to the existing FC protocol (FCP). Currently a project under ANSI's T11 group, the FC-SATA protocol would simply add the SATA Tunneling Protocol (STP) to FCP, just as SAS does today. Brencic expects project members to deliver a first draft of the standard to ANSI in early 2006.
Migrating old files curbs disk costs with the help of automated file migration software--the cornerstone of information lifecycle management (ILM)--IT shops are finding they can go much longer before having to add disk capacity to their primary storage systems.
"We found that in the first 18 months [of putting in place an ILM strategy], we didn't have to upgrade our primary storage once," reports Chris Johnson, network administrator at Mustang Engineering L.P. in Houston. A Network Appliance shop, the firm is using File Lifecycle Manager from NuView to migrate inactive files from high-end filers to an ATA-based NearStore. So far, Mustang has migrated approximately 6 million files, totaling 4TB, to NearStore.
Similarly, Quicken Loans, a Livonia, MI-based online home financing lender, recently celebrated its one-year anniversary of using EMC Legato DiskXtender plus EMC ControlCenter (ECC) to migrate older files. In that year, the firm didn't need to add capacity to its primary storage tier. "Because we move files off the high-performance disk, we recoup that space," says Scott Roemmele, SAN engineer team leader.
After a year of monitoring its storage usage and developing file migration policies using Arkivio's auto-stor, the federal manufacturing and technologies division of Honeywell expects to recoup approximately 80% of the Hitachi Thunder and Lightning SAN array capacity assigned to unstructured file data. Approximately 5TB of those files will be migrated to an EMC Centera content-addressed storage (CAS) device. "Unstructured data makes up the majority of our data," says Matt Decker, the division's lead systems engineer. "This will help a lot."
These examples are fairly typical of environments that have deployed automated file migration software. File migration software "forces you to have better discipline" about what types of files deserve to live on what class of storage, says Tony Asaro, senior analyst at Enterprise Storage Group (ESG), Milford, MA.
For example, Quicken Loans spreads the 6.5TB of file data it has under management across three basic tiers: new files and files that have been accessed in the last 90 days reside on a primary Symmetrix array, while documents that have been accessed between 90 and 180 days go on a Clariion with ATA drives. Any files that haven't been accessed in 180 days go into deep archive storage on an EMC Centera. Quicken's internal customers, meanwhile, are blissfully unaware of where a file resides--they just see a stub. "This has allowed us to keep a lot more data online," says Roemmele.
Before putting DiskXtender and ECC in place, Quicken Loans put the vast majority of its files on Symmetrix and just left them there. "We were just letting the beast grow and grow and grow," says Roemmele. Eventually, "we came to the realization that we needed to buy less-expensive disk."
ESG's Asaro claims that even with the additional cost of file migration software, you'll still save money by moving data from tier-one to tier-two disk, which is priced at approximately 60% of primary storage.
However, if you're architecting in a new tier of storage, that may translate into some new upfront costs. "I always say that ILM is a strategy," says Honeywell's Decker. "It's a little bit expensive up front, but down the road, it's the most intelligent thing you can do." Otherwise, he says, "you're going to suffer with compliance issues and silly spending on additional resources."
According to Buzz Walker, VP of marketing and business development at Arkivio, environments that see the most benefit from migrating data off primary storage have a minimum of 3TB to 5TB of unstructured primary data --"after that point the benefits are just exponential," he says.
But environments with less data can also save big by clearing out old data, Walker says, notably around backup optimization. "That's where the cost justification is the strongest," he adds.
When it comes to inactive files, "you already have one good backup, why do you need another?" Walker asks. If those inactive files are moved to a second tier of disk, IT administrators can recalibrate their backup policies and back up less frequently, thereby saving money on tape media, drives and offsite storage. It goes without saying that completing your backups within your window suddenly gets much easier. "Now, on my primary storage, I can hit my windows whereas I couldn't before," says Walker.
Meanwhile, storage vendors are bracing themselves for sales of primary disk to be affected, says Asaro. "It's an eventuality they understand and are preparing for." But there may be a silver lining: What vendors lose in the short term to cheap storage sales, they may make up in the long run by locking customers into their storage brands.
"That's the beauty of [a product like EMC's] Centera," says Asaro. "We used to worry about [a] three-year lock-in, but with retention times of seven years or 30 years, now you're locked in for that much longer."