On Wed, Jun 2, 2010 at 3:31 PM, John Robinson <john.robinson@xxxxxxxxxxxxxxxx> wrote: > > I'd consider a larger than default chunk size for the RAID-5. +1 absolutely When building a server for storing mostly video files on RAID-6, the files being around 500MB to 4GB in size, I ended up with 256KB chunks. Gained quite a bit of speed compared to the default! (Also remember to play around with stripe cache after you're up and running.) (I did some very basic benchmarking. I repeated some tests I found online, which indicated that RAID chunks around 256KB were fastest. My tests agreed.) I would consider not building a swap partition at all, but rather use a swap *file*, that you can put on any partition. Here's a howto: http://www.linux.com/archive/feature/113956 Alternately consider putting the swap file on a RAID0 for the most speed, which also costs you a smaller chunk of each disk for the same swap size. The safest swap is mirrored though, and I'd agree with RAID10,f2 in that case. Hmm ... no conclusion ... I'm thinking out loud more than I intended to. To state something concrete: I think my conclusion du jour regarding swap RAID is probably that swap space isn't that important anyway for regular use - one should always have enough memory to begin with so relegating swap mostly to emergency use. Regarding swap as an emergency feature, it's probably best to have it mirrored - you wouldn't want your swap space to ever vanish in the middle of anything. And so I'd go with RAID10,f2, which is by far the fastest mirror. -- Kristleifur -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
