The number of inputs and outputs per second from your networked storage can dictate the success or failure of your whole architecture. Danny Bradbury interviews systems administrator Trevor Pott and finds out how networked storage can be configured to avoid bottlenecks and maximise performance. Watch Video var MediaPlayerVars …
Not that great an interview
The interview is titled "How to get the best from your IOPs" .. but doesn;t actually get onto that topic .. was rather looking forward to something interesting.
If it could be expanded into which of the technologies available are best for a given scenario (eg. technology to use for a datawarehouse, vs a billing system) then it would be quite informative.
agree with above comment
"stick with what you already use" sounds a bit dogmatic to me.
If you're consolidating a decent number of servers (some with big application workloads) on to a small number of virtualised boxes, sticking with iSCSI might not be the best bet - making the decision based on circumstances and requirements tends to be the best bet!
I also loved the comment about the overhead on FC - does anyone drive 10GbE at "rated speed" in the real world? Overhead / inefficiency on Ethernet is a bigger issue than on FC!
I wasn't going to mention that "overhead" comment the guy made,as I wasn't sure where he got it from, FC started as a nice tight transport protocol with little overhead (way better than Gig-e when FC first came out), I'd imagine 10GbE is better designed but wonder by how much ? ... will have to go read and satisfy my curiosity :-)