In 2010, RAID disk arrays still show up in the small- to medium-sized business (SMB) and prosumer market. In fact, according to analysts, when it comes to RAID purchases, the decision points are often rising above the level of RAID itself to strategy, architecture and new device choices.
RAID purchasing trends
Roger Cox, research vice president at Gartner, said he sees three big trends starting to impact those thinking about purchasing RAID. One is the advent of various storage schemes that add bells and whistles to the RAID formula such as mirroring layered on top of a traditional striping of data. But few of the additions in spinning disk capability impress him as much as the potential for marrying solid-state storage to disk -- producing ultra-fast IO while retaining the more favorable cost structure of traditional RAID.
The second trend is the cloud data storage and cloud data backup, which Cox predicts will see significantly more traction for enterprise data storage in 2010. For him, the cost and simplicity arguments make a lot of sense, meaning more and more companies will simply want to plug into storage located somewhere else, eliminating decision points about RAID purchases.
Third, there is the effort by vendors such as Cisco Systems Inc., Dell Inc., Fujitsu, Hewlett-Packard (HP) Co. and IBM Corp. to offer devices that provide server, storage and networking functionality in one integrated package. "They are trying to convince the user market that this is the way to simplify and reduce cost and complexity," he said.
However, noted Cox, all of these trends are either still held back by market conditions or the maturing of technology.
How do you define RAID?
Perhaps that's why others see the future of RAID looking more like the past. "First and foremost it is a myth that traditional RAID is dead, that's hype," warned Greg Schulz, founder and senior analyst at the StorageIO group. "RAID is relevant because it has become so transparent you don't even realize it is running on your desktop or on a desk-side NAS [network-attached storage] system," he said.
Schulz said what is happening is that traditional RAID products are just gaining more functionality, becoming easier and more transparent, so you don't have to worry about details like chunks and strips. "It is increasingly a set-and-forget process," he added.
Among the vendors with offerings that are recognizably RAID but incorporate some extra features are 3PAR, Compellent, EqualLogic (now owned by Dell), Isilon, and LeftHand Networks (now owned by HP). The so-called post-RAID vendors/products include Cleversafe dsNet, Data Robotics Inc. (Drobo), IBM and its XIV Storage System, EMC Corp.'s Atmos and Centera, Panasas ActiveStore, Permabit's RAIN--EC and Seanodes Exanode (which virtualizes storage assets to convert unused internal disks and DAS into a shared storage array), among others.
Schulz said it comes down to your definition of RAID. "The basic notion is still multiple drives with some extra protection and if you want to call it RAID, good, if not, God bless you," he said.
For SMBs fretting over what to buy, Schulz said, to take a step back, consider ways to create a flexible solution perhaps with RAID 1 mirroring for a high-performance database; maybe RAID 5 for a NAS file system and some bulk data protection with RAID 6 to reduce costs.
RAID options for SMBs
"Plenty of products exist, especially for SMBs, because the upper end of the SMB space is the bottom end of the enterprise space and the low end of the SMB space is near the upper part of the SOHO market," he said. Finally, regarding the many new permutations of RAID -- such as Data Robotics DroboPro or IBM Corp.'s XIV Storage System -- Schulz said it is important to remember that so far they represent only a tiny portion of the storage market.
Like Schulz, David Hill of the Mesabi Group also sees traditional RAID continuing to play a vital role, with some limits. For instance, although modern disk arrays can handle many RAID groups, Hill says each RAID group is a logical collection of disk devices and has to be managed by the group. "When scaling to petabyte scale, this may not be the appropriate approach," he said. Still, whether at the file, block, sub-block, or byte level, one or more extra copies of the data need to be retained to meet the availability requirements users want.
This redundancy at the data level means that the data could be spread over an array using algorithms and/or heuristics that can dynamically move the data as necessary, for example, when additional disk capacity is added to the array. "If the vendor wants to make sure that a minimum of three data protection copies are kept in addition to the working production copy, enough space has to be reserved to do that. That can absorb a lot of disk capacity. Hill says some of that "wasted" space can be avoided by employing data deduplication and compression. But by definition, "somewhere, somehow four copies -- one production and three data protection copies -- must be found in the array and each copy must be physically isolated from each other so that no two copies are affected by any one disk failure," he said.
Finally, when Hill looks at the companies touting "post-RAID" data protection solutions, he worries about one overarching issue. "Traditional RAID had the benefit of being a standard approach; these other approaches are likely to be vendor-specific. That may not be a bad thing -- as long as you are confident that the vendor has got it right," added Hill.
About this author: Alan Earls is a Boston-area freelance writer focused on business and technology, particularly data storage.