What Storage System? It Depends.

When at a storage event or even responding to comments, every so often a specific question comes up about which protocol, drive type, backup application should a user implement for their specific environment. The answer that should come back most often is "it depends" because that's the truth.The reality is that for most situations, you can get almost anything to work. Its a matter of how hard you have to work at it.

For example, if you are setting up your first virtualization project, you can get it going and get two hosts talking with the same amount of ease/difficulty on iSCSI as you can on NAS, AoE, SAS or...dare I say it?...Fibre Channel. The test for any of these environments is how they change as the environment scales? When you have 50 hosts with 500 virtual machines, the complexities of these different protocol options and the storage systems that support them become important factors in the amount of time spent managing storage. Conversely, if you only have $5k to spend, all the amazing things that a $50k SAN does really doesn't matter, you're going to have to make that $5k solution work.

Beyond the obvious budget issues, it just depends on how far you are going to take whatever project you are working on. In my experience, the greater the demand for either performance, scale or longevity, the more important it is to understand the differences between these technologies. Storage systems and protocols that can address more needs are almost always the more expensive solutions. Something that is cheap and easy to install may not be easy to operate when 500 virtual machines are pounding on it.

I should put in the obligatory suggestion that after you have read as much as you can and have asked a lot of questions, you eventually have to do your own testing. Labs can be as much of a pain to keep together as production systems sometimes, and they still don't tell you everything that will happen in production. Also, I'm sensitive to the reality that as a user, you are stretched thin as ever. The time to properly evaluate the top three or four vendors in a space just doesn't exist any more.

Lately we've advised the vendors to expect very long test cycles. You can only test so much in 30 days, and it is really not indicative of how the system is working. Ask me about a system that we have had in the lab for three or four months is working, and I can give a much more confident assessment. You can take comfort in the fact that most storage systems can be made to work. The amount of research and investigative work you do is the key to how difficult it will be.