Published: 10 Feb 2006
A toaster oven in the data center
Vendors may claim midrange arrays rival enterprise systems feature for feature, but user skepticism threatens to take some of the air out of the hot midrange market.
MY TOASTER OVEN will cook a turkey--at least that's what the manufacturer claims. It even came with detailed directions describing how to roast the big bird. According to the specs on the box, the only difference between this little countertop marvel and the conventional oven sitting next to it, apart from physical size and price, is that the big guy requires a special three-phase power outlet. In the storage environment, an analogous situation exists between midrange and enterprise storage arrays, right down to the special electrical outlets.
Why does anyone buy a conventional oven or an enterprise storage array? It turns out that folks in general are pretty smart when it comes to the old adage of "you get what you pay for," and they're rapidly becoming disenchanted with midrange storage despite the more modest price tags. Vendors have been adding barely functional features to these systems while their sales forces aggressively push these little wonders. But skepticism abounds in the user community, and virtualization might just stick a fork in today's hot midrange storage market.
The right tool
Midrange modular storage devices have had a profound impact on the enterprise in the last few years. EMC's 1999 purchase of Data General showed that an alternative to monolithic storage was recognized at all levels, and less-expensive modular storage systems soon began to appear in corporate data centers. The appealing concept of tiered storage showed the illogic of placing all of a corporation's information on expensive, high-end systems; it was then easy to make a case to introduce a reliable, high-performance, low-featured storage system as an alternative to the high-end hardware.
But vendors weren't content to simply sell bulk storage. Instead, they quickly began to add high-end features like array-to-array replication, internal mirroring and snapshots to these low-end systems. Buyers could be forgiven for thinking there was little difference between midrange and high-end systems apart from cost. I've even heard salespeople claiming that their company's high-end and midrange storage lines are essentially similar.
But experience has shown that relying on midrange systems for high-end applications is akin to relying on a toaster oven to prepare your Thanksgiving dinner. It might work, but most of the time it doesn't. It always pays to use the right tool for the job, and these systems are far more limited than they may appear to be at first glance. Many midrange arrays have absolute restrictions on the number of logical unit numbers that can be mirrored or replicated, and most are incapable of load balancing I/O across the two controllers they're uniformly equipped with.
Users have also reported reliability issues with midrange systems, especially with regard to widely used low-cost ATA disk trays. And post-sales support for these midrange products may sometimes be handled by third parties, a situation that may not always be understood beforehand. Difficulty with management and reporting software, and a general lack of familiarity are other common complaints. One company I spoke with still hasn't figured out how to automatically import utilization statistics from its midrange system to its standard corporate reports.
Bulk to basics?
Many companies have decided to abandon the advanced features of midrange storage, relegating the boxes to be used for bulk disk space. In many shops, you'll find a lot of midrange storage sitting behind NAS appliances or file servers, backup systems, virtual or blade servers, or even virtualization devices. Although it's likely that vendors didn't envision their midrange systems being used in these configurations, each of these applications maximizes the utilization of low-cost, high-performance disk devices without the troubles that come from using the wrong tool for an enterprise storage job.
These alternative products often come from established storage heavyweights. IBM's SAN Volume Controller and EMC's Invista promise to take the smarts out of the storage array at the same time these vendors are scrambling to prove how smart their arrays are. Only Hitachi Data Systems' (HDS) TagmaStore sticks with the traditional array-centric approach. NAS heads, which use ordinary Fibre Channel storage rather than proprietary disk shelves, are gaining widespread acceptance. Many of EMC's other products, including Legato NetWorker, the Legato Extenders and VMware, are also heavy users of basic bulk storage.
But the obvious next question is, why buy midrange storage at all if none of the bells and whistles it comes with will be used? With the possible exception of support concerns, bulk storage from second-tier vendors is just as viable as second-tier products from the biggies of the industry, and more cost-effective as well. And some major server vendors, including Apple, Dell, Hewlett-Packard (HP) and IBM, are now offering interesting bulk FC storage devices.
So why run VMware on an EMC Clariion? Why use IBM's DS4000 behind SAN Volume Controller? If you don't plan on using any of the more advanced features that are offered with those midrange systems, why not stock your data center with Apple's Xserve RAID or HP's MSA instead? The Apple device has essentially no features, so you don't have to worry about paying for features you won't use.
The next revolution
Perhaps we're entering a new era in networked storage, where the feisty midrange storage systems that have been pushed so vigorously by aggressive vendors are wiped from the enterprise arena. Amazingly enough, it may be those same vendors who kill off modular arrays by cramming too many features into them while offering new devices and technologies--like virtualization--that eliminate the need for them.
Look at the up-and-coming products in the market today and you'll see grid-type devices that gang together low-end and low-cost storage to offer high-end features. Many next-generation iSCSI arrays and NAS boxes use ATA disks and Intel CPUs in large, scalable clusters to offer data protection, replication and redundancy on the cheap. Again, the big vendors are playing in this market, and next year's acquisitions of grid-storage upstarts by the big guys will probably solidify grid storage as a new tier. I'm not sure how any vendor pushing virtualization and grid storage can continue to sell mid-priced storage arrays with middling functionality.
High-end storage systems will always be needed in the enterprise, and monolithic storage arrays are likely to fade from the landscape only when virtualization is everywhere. But more than one company is wising up and seeing the once-obvious benefits of adding a modular storage system from a big-name vendor vanish. The cost of additional management software and staff training is colliding with the falling prices of enterprise storage; reality checks on the claimed levels of functionality and reliability are also helping to take the bloom off the midrange storage rose.
In addition, future technology developments don't look too positive for midrange storage arrays. Although smaller companies may want to move up to an EMC Clariion or an HDS Thunder, the next wave of grid storage systems will likely offer far better price points for entry, seamless scalability and functional enterprise-class features.
So say goodbye to midrange storage arrays as we know them today. These half-baked turkeys will be cleared off the data center floor unless storage vendors can come up with a new cost/benefit justification.