By Jim Damoulakis, ComputerWorld.com,
As IT infrastructure budgets contract with the general economic slowdown, an oft repeated refrain within large datacenters is the desire to cut storage spending to near zero. The general idea is that there is so much over allocated storage that it should be possible to recover and reallocate that storage before buying more. While there are likely some opportunities to do this (see " Opinion: Finding a home for orphaned storage ), the overall exercise is a bit like trying to eliminate waste in the Federal government - it's easy to find, but eliminating it is next to impossible. In fact, in some ways it's more complicated since we're not trying to eliminate capacity, we're trying to recapture and then reallocate it - a process fraught with challenges.
As a result, one must seek out other opportunities for economizing within the infrastructure. An area that I believe is ripe for re-evaluation is the knee-jerk tendency of large organizations to default to Fibre Channel for storage connectivity. While fibre channel has become a reliable technology and continues to get faster, it is an expensive and complex method for connecting to storage. Despite IT reluctance to not fix what isn't broken, if cost reduction is this year's rallying cry, it's important to selectively consider alternatives.
If we consider one area currently experiencing significant growth - VMware deployments, the common design practice within larger organizations is to deploy on a Fibre Channel infrastructure, while smaller organizations (who are less likely to have invested in fibre channel) favor iSCSI connectivity, presumably based on cost. (The entry costs to Fibre Channel can be high not just in terms of equipment, but in developing a management and support capability, as well).
When the prospect is raised about using an alternative to Fibre Channel, the twin specters of performance and reliability are often raised. Unfortunately there is often not a whole lot that is understood about what the actual performance needs are. There also haven't been many benchmarks available. One exception is a study earlier this year from VMware that found that Fibre Channel did indeed offer the highest bandwidth when compared with hardware iSCSI, software iSCSI, and NFS. However, the Fibre-Channel advantage appears to be largely based on the fundamental wire speed advantage (4Gb versus 1Gb) over the other protocols - an advantage that could potentially be erased by 10Gb Ethernet (pending VMware support for all protocols - currently hardware iSCSI is supported). Furthermore, it found that the other three protocols by-and-large performed similarly, and it concluded that for the right application mix, the alternatives could offer a significant cost-performance advantage.
The dark horse in this picture could be NFS. Large organizations have been deploying NFS for 20+ years, and, generally, it enjoys a significant advantage over block-based protocols in terms of manageability. The fact that this ease of management may be had without a substantial performance penalty should cause organizations to give it another look.
Now, let me stress that one benchmark does not a best practice make, and I'm not suggesting wholesale replacement of Fibre Channel, but for specific application categories, there are compelling lower-cost alternatives. As always, there should be appropriate testing and proof-of-concept trials, but in a time of "zero" budgets, it behooves us to reconsider our fundamental design biases.
Jim Damoulakis is chief technology officer of GlassHouse Technologies Inc. , a leading provider of independent storage services. He can be reached at jimd@glasshouse.com.
Copyright © 2008 IDG. All rights reserved.