In the virtualization community, one of the most frequently debated topics is FC vs. Network-based storage: which is better? Some IT professionals have deeply rooted convictions, some say they don’t have a preference, and some give that age-old consultant response: it depends.
It’s not uncommon to hear that, when performance is a design requirement, FibreChannel is the only way to go. Don’t even consider network-based storage when high-performance is required. Of course, you’ll need to define what “performance” actually means in terms of IOPS, latency, throughput, et cetera. But FC still always wins, right?
You’ve probably noticed that the virtualization world is undergoing a sea change with regard to storage. Lots of attention is paid to companies like PernixData, who has an innovative solution for server-side flash. PernixData promises to de-couple storage performance from storage capacity. With storage moving ever closer to compute, at some point you’ll find yourself asking: do we really need to design for performance at the SAN / NAS layer?
Suddenly, server-side flash satisfies your performance requirements. Now the question becomes: do you need to maintain a separate physical network for a single protocol? If you’re doing server-side flash, you probably have a 10/20/40 (maybe 100?) GbE network, and can accomodate the throughput for storage traffic while still meeting bandwidth requirements for the rest of your data center traffic.
I suspect that we’ll see a migration from FC to network-based storage once adoption of server-side flash ramps up. Storage performance will dramatically increase, data center cabling and complexity will decrease, and the importantance of QoS and monitoring will increase. Exciting times for network, storage, and virtualization engineers!