Other media (flash and optical) which do not have the benefit of Enterprise lineage currently boast even more dismal statistics.

If things don’t start to change, this issue will come to the forefront a bit sooner than most folks realize. The sky is certainly not falling this year, or next, but given exponential consumption and capability trends….

Eric, during an “average week” (or whatever your granularity of uptime is), can you send me the amount of “data read” shown by Activity Monitor. It’s a global that doesn’t provide per device information, but it’s a good way to get a feel for where you currently stand.

Those of you who do a lot of I/O and do NOT see this as an issue, ask me for the CERN report or simply try it out for yourself. Create a file with a well known pattern, read it back, repeat. There are some fine details (throttling your I/O, etc.) but the workload is simple and common. Try it on your favorite media.

That WiebeTech looks interesting. I would be curious to know how well it works over time. What kind of speeds are you getting writing to it?

I’d like to see application benchmarks like FCP once they’re configured to properly use the Speed and Data partitions. I’m not convinced that the modest and occasional performance benefits of the Speed partition compensate for the hard partitioning and forced seeking that results. I could easily see overall performance being degraded; in fact I’d expect it. “Speed” partitions should be on dedicated, very fast drives.

I’m disappointed that RAID 10 on OS X doesn’t match RAID 0 read speeds. Perhaps it does and the benchmarks used simply can’t show it. Given sufficient command queuing (not automatic with desktop systems and benchmarks!), identical drive counts should yield similar read speeds for RAID 0 and 1 or 10 unless the programmers are stupid.