Look at Duplication Technologies
Networked storage capacity in the Fortune 1000 will grow a projected 24 percent in 2011, according to a May storage market study conducted by TheInfoPro, a division of The 451 Group. It’s easy to see why: Increased virtualization, larger data sets and more complex applications all mean that managing storage is still high on IT's to-do list.
Storage area networks (SANs) are established technologies, and many IT managers have learned lessons from the school of hard knocks. Some of them are on third- or fourth-generation equipment after more than a decade of use.
Let’s examine the best practices for managing your SANs, how to buy the right equipment for your needs, and how to implement the best disaster recovery, storage virtualization, storage tiers and clustering options.
• Find and stick with a good storage management solution. These are products that can help assign storage to particular servers, set up redundant drives or clustering, handle continuous data protection and other advanced functions. These solutions are worth the initial cost and trouble to implement because they can save time down the road when you need to make changes to your storage infrastructure.
Bryan Peterson, associate director of technical services for Utah Education Network in Salt Lake City, runs a statewide network that connects all public K-12 and higher education campuses. The network provides central IT services for various education courseware applications and back-office purposes.
The organization purchased DataCore Software’s SAN Symphony for managing its virtualized storage network. “This allows us to be array-agnostic, since we manage everything with DataCore,” Peterson explains. “We would like to stick with one array vendor, but we are price-sensitive since we are a state agency. This makes it a lot easier. Plus, we can attach all different kinds of storage to our network and manage it centrally.”
Strand Associates, a consulting engineering firm headquartered in Madison, Wis., switched to Dell Compellent storage arrays because of their built-in management tools and ease of use. “We like the fact that Compellent comes with thin provisioning and tools that can automatically manage where the data resides on our disks via automatic tiering and systemwide striping,” says Justin Bell, a network engineer. “You can be really hands-off with the process, and it takes just two or three clicks to expand a drive. This is light years from where we were before.”
• Look at deduplication technologies. Deduplication is built into a variety of SAN and backup products, and it can offer big benefits in terms of storage costs and backup windows.
Lovell Hopper, manager of the Technology Services Branch at the California Emergency Management Agency in Sacramento, uses IBM’s Tivoli Storage Manager to handle 35TB of usable storage. The agency started several years ago with 200 physical servers, but it has reduced this collection by using 80 virtual machines running on 20 Hewlett-Packard blade servers. “It is cheaper to buy a new blade than to buy the additional RAM for an older server,” Hopper says. “Plus, we can purchase higher density and faster CPUs with faster storage fabric interfaces.”
The agency cut its backup times from hours to minutes by using Data Domain’s backup solution. They were able to cut storage needs by a ratio of 50:1. “We have lots of geospatial data, and we saved tons of space when we performed dedupes on the backups,” says Hopper.
Utah Education Network had a similar result. “We had 16-hour backup times with our Oracle database, but with direct-to-disk and deduplication, this dropped to four hours,” says Peterson.
Strand’s Bell says it’s important “to have a way to manage your bandwidth usage if you’re going to be doing replication across a wide-area network or lower speed connections.” He purchased FalconStor backup technology to minimize the amount of data that’s replicated down to the sub-block level. “It’s very efficient, and we see about 25 percent of the traffic going across our WAN as a result of this technology,” he adds.
• Thin provisioning is in the thick of things. You can create a lot of wasted space when you set up a new SAN array, but thin-provisioning products such as Symantec/Veritas Storage Foundation Manager, XenServer Essentials and Dell/3Par can cut that down substantially. Here’s how:
When you provision your SAN, you generally don’t know exactly how much storage you’ll need. This often makes you err on the high side with volumes large enough to meet your requirement for the life of the server. The same thing happens when you create the individual virtual machines on each virtual disk partition.
“We saw a lot of value with thin provisioning,” says Peterson of Utah Education Network. “We have oversubscribed our storage by about 20:1, and we don’t have any issues. We can increase SAN disk utilization to 60 to 80 percent.
“Without thin provisioning, we would have had to come up with really good storage guesses and spend more time managing the storage and purchasing a lot more disk drives. Thin provisioning gives us a fairly good life-cycle-management process, and we can easily move data between storage tiers.”
• Redundancy isn’t what it used to be. “Redundancy can be your friend in times of failure,” Peterson notes, “but, if taken to an extreme, it can make it hard to find unexpected failures. Make something only as complex as it needs to be.”
• Document everything. “You can’t document enough of your storage network,” he says. “This is important: Nothing is worse than going to one of your servers and finding unlabeled cables coming out of it and not knowing where they go.”
• Don’t cut corners. In the City of Altamonte Springs, Fla., Lawrence DiGioia, information services director, runs twin Xiotech SANs with 50TB of storage in each. To provide additional protection to systems from hurricanes and other storms, the city renovated a 770,000-gallon ground-level water tank with eight inches of reinforced concrete, and integrated it into the main data center.
“It was very cost-effective and green too,” DiGioia says. “The biggest mistake I’ve seen is that people get disappointed in a SAN’s performance because they cut corners, didn’t invest in training, or didn’t bring all their business partners together in one room to plan out their infrastructure to ensure success. Spend the money up front, do proper planning, engage your business partners and provide some training.”
This is especially the case for buying the right amount of backup. “Recovery-time objective is just as important as recovery-point objective,” says Strand’s Bell. “It doesn’t matter if you have up-to-the-minute backups if it’s going to take you a week to restore them. If you have a second backup copy on-site, you can be back online in a couple of minutes.”
Bell learned that the hard way several years ago when a hardware failure took a server down for several hours.
• Flash the cache. Solid-state drives have come into favor for applications in which performance is more important than cost. Storage vendors are looking at ways to marry the high performance of solid-data memory with the lower cost of rotating media to provide the best possible storage solutions.
Vendors include EMC Fast, FusionIO, Viking, Violin and WhipTail Technologies. EMC has shipped some 14 petabytes (that’s 14,000 TB) of solid-state arrays on its own. A solid-state solution is often part of a collection of tiered storage devices, whereby information can move from one device to faster rotating drives on a SAN to slower near-line drives for backup.
“Hybrid solid-state/rotating disk solutions are inevitable,” says DiGioia of Altamonte Springs. “We are looking [forward] to virtualizing and booting our desktops from the SAN.”
• Toss tapes. Given the continuing price decreases for disk storage, tapes don’t make much sense. “Tape management is obsolete,” says Ravi Sachidanandam, an assistant professor with the Mt. Sinai School of Medicine, in New York City. He has several hundred terabytes of SAN storage to handle a variety of advanced medical applications, such as gene-sequencing machines.
“Each week, these alone generate a terabyte of data, and they are running constantly,” says Sachidanandam. “When we last got some quotes for tape solutions, they cost almost as much as hard drives, so it didn’t make any sense.”
“Get rid of tape,” advises Utah Education Network’s Peterson. “We have gone completely to disk, even for backups. The hassle factor of tapes isn’t worth it. Virtualize your backups and use real disks. While the capital costs are initially hard to swallow, the disk backups really perform, and with dedupe, this setup can bring costs way down.”
Altamonte Springs’ DiGioia agrees: “Tape is cheaper, but it’s only 70 percent as reliable as disk backups. Disk-to-disk might cost more money initially, but consider the value of the data you’re backing up.”