Re: To RAID or not to RAID (...or how to RAID)

Anton Buijs <aammbuijs_at_xs4all.nl> wrote:
>I agree we should set up a benchmark, but that takes time and -as Niall
>writes - what is a good benchmark?

One that doesn't have me sitting in the benchmark centre after 9pm for the third night in a row... (sorry - just venting).

On topic, I've had no problem with SAME over the course of many benchmarks. As was mentioned, various TPC-C benchmarks use SAME, and you can bet the hardware vendors do everything they can to get the best performance on such a high-profile test. Obviously, your mileage may vary.

Some hardware vendors are now even claiming RAID 5+0 (striped RAID-5 sets) can offer adequate high-end performance. Obviously gives more spindles for the money than RAID 1+0, but with a serious performance impact when you have a drive fail, as everything has to go through the degraded controller.

-- 
Andrew Mobbs - http://www.chiark.greenend.org.uk/~andrewm/