2011-05-10, 02:27
IAmNotAUser Wrote:To add to Superob's post, a cache drive is recommend to be the larger than the amount of data you would aim to transfer for a 24 hour period but it doesn't have to be any particular size in comparison to your other drives. My Unraid build currently has a 2TB parity drive, 2x 2TB data drives, 2x 1TB data drives, 1x 500GB data drive and a 500GB cache drive, with the automated mover script occurring twice daily (as any data that is somehow lost whilst on the cache drive isn't protected).
EDIT: That said, it seems in your test you didn't have a parity drive either, which renders this whole discussion about parity calc being the bottleneck moot. If you had a parity drive and a data drive only for your tests, then that's the cause of your slower speed. If not, then I don't know.
I was thinking more about the reasoning behind why the transfer speeds may bet a little slower with unraid. Once big thing that comes into play that I hadn't thought about is the actual read and write speed of a hard drive. A single 7200 rpm hard drive has a write speed of about 50-60 MB/s which is definitely where the bottleneck was on my test unraid system. Furthermore adding a cache disk wouldn't speed up transfers in my test setup but I could certainly see how it would speed up an environment that employs parity (mosts users here). I think one solution that may benefit anyone experiencing slower transfer speeds would be to make your cache drive a smaller (say 250gb) 10k raptor drive that has a faster r/w time which would allow you to utilize the speed increase of lan (125 MB/s MAX). I don't think anyone here transfers more than 250gb in a 24 hour period. I may have a raptor drive laying around, if I do I will clock it and let you all know how it tests.