A multiprotocol array is a classic "black box" solution, one designed to solve the types of storage problems that arise when a company needs to support both SAN and NAS in their storage infrastructure.
Storage managers have been eagerly buying into the simplification offered by multiprotocol arrays. Why purchase -- and worry about -- two (or more) arrays needed to support different kinds of block- or file-based storage, when one multiprotocol array can do the job?
Another inherent advantage of multiprotocol arrays (also known as unified storage or hybrid storage) is flexibility. Since they are, to varying degrees, protocol-independent, they allow storage administrators to work with a wider range of products such as switches and servers.
Likewise, the simplicity of combining functions in a single box means that multiprotocol arrays can reduce the amount of cables and can put more kinds of storage within a single management framework. Multiprotocol arrays may even save you money, since they can consolidate storage and, thus, have the potential to increase storage utilization.
However, money savings are no guarantee. What's more, in some high-end applications, such as supporting large databases, they can't necessarily match the speed of traditional single-protocol arrays.
The procurement process involves sorting through the features available on specific products. For example, many multiprotocol arrays include snapshot,
Combining SAN and NAS
Even the definition of multiprotocol array can be a bit slippery. However, storage industry observers tend to define multiprotocol array as one that combines SAN (or block-based storage) with NAS (or file-based storage). Native block-based storage includes three protocols -- SCSI, iSCSI, and Fibre Channel (FC) – so arrays that can handle this SAN subset are occasionally referred to as multiprotocol, too. NAS systems, for their part, eliminate the need for a separate file server, instead incorporating file server protocols such as Network File System (NFS) and Common Internet File System (CIFS) into the storage itself.
The history of the multiprotocol array begins when NAS vendor NetApp (then Network Appliance Inc.), got tired of seeing so much potential business go to rival SAN vendors and made the momentous decision to try to adapt its NAS box, which was oriented toward handling files, into an omnivore that could also handle block-based SAN traffic. The product that resulted, the FAS900 series, rocked the industry and attracted the attention of both traditional storage vendors like EMC and newer and smaller NAS vendors. Today, the market for multiprotocol arrays includes companies offering NAS-rooted products (NetApp being the prime example), and those, such as EMC, with SAN-rooted products.
However, says Bob Passmore, research vice president at Gartner Inc., "while almost anyone can build a unified device on paper, it is more work to do it properly." Even today, history and lineage have a lot to do with what makes multiprotocol arrays more or less successful.
NetApp sets the pace in mulitiprotocol arrays
Both Passmore and Arun Taneja, principal and founder of the consulting firm Taneja Group, credit NetApp with setting the pace and point out that NetApp's ability to add block storage access was helped by development of its Write Anywhere File Layout (WAFL) file-system layer. To make all the storage elements play effectively together, NetApp also developed a dedicated operating system called Data ONTAP which has gradually been expanded to support not only NFS and CIFS but also iSCSI and Fibre Channel. (There are currently two variations of Data ONTAP: the 7G and GX versions. The latter incorporates virtualization technology.)
In contrast to NetApp, EMC got into the multiprotocol arena by adapting its line of SAN products. EMC had Clariion and Symmetrix which started off as FC only. When iSCSI emerged, that became an alternative configuration choice on Clarion – and ultimately the products were offered with both FC and iSCSI. EMC introduced its first multi-protocol system (unified block and file) in March 2006 with the Celerra NS350.
Taneja says the current NetApp lineup fits the needs of a lot of organizations and is flexible in that it can support consolidated file services or database applications, while individual servers with NICs can access additional NetApp storage when needed.
Over time, EMC has expanded its lineup and has added many of the tools and capabilities first provided by NetApp. Passmore says there are a number of other products that offer some kind of "unified" storage but at the tradeoff of complexity. For instance, EMC has simply taken a FC box and put another server – a NAS head --"in front" to direct storage traffic based on protocol.
Passmore says he believes the NetApp approach is especially appealing since it provides a single software stack for all its products, from those geared to SMBs to those aimed at the enterprise, versus the vendor approach that positions different products with different technology for different parts of the market.
According to Taneja, part of what makes NetApp work well for snapshots is the fact that it is a "write behind" approach. "They never write over anything, except for a garbage collection function and they distribute data randomly across disks, which helps with load balancing," he said. "In the case of conducting a snapshot, it also means that data does not have to be moved."
Other vendors use the snapshot as an initial marker, but when IO operations are conducted, the arrays must actually pick up the old data and move it someplace new so they can overwrite the location with incoming array, all of which absorbs a lot of processing cycles. "For NetApp, a snapshot is a non-event that doesn't cause any problems downstream," says Taneja. The same architecture also enables much faster backup when using an array as virtual tape.
Are multiprotocol arrays the best choice for every scenario? Probably not. For instance, multiprotocol arrays can sometimes be challenged when NAS and SAN IO requirements compete, for example in accessing large files, copying large numbers of files or performing backups. For example, in the case of NetApp, where NAS and SAN pass file- and block-based protocols through the NetApp WAFL file-system layer, the number of CPU cycles required can slow access significantly.
"If you want absolute performance on one side or the other, you probably don't want a multiprotocol array," admits Taneja. On the other hand, for most situations, the inherent compromises of multiprotocol deliver good performance and great value.
About the author:Alan R. Earls is a Boston-area writer focusing on the intersection of technology and business.
This was first published in June 2008