The 03.07.20 at 18:05, richard wrote:
From personal experience, but without using a stopwatch and comparing two separate installations, RAID (with apologies to Philipp for abbreviating this from "Not Really Real RAID"), I would only say it seems not to be slower. It is clearly not the difference in getting a new MB/CPU and cetera. Test results I had seen earlier (Anandtech, maybe?) suggested a performance increase of maybe 10% if you were lucky in striped mode, maybe a couple of % if anything in mirrored mode, though reads were a little quicker than writes. Personally, I do not see (Philipp's hated) RAID as a solution to performance issues, but rather as a solution to other issues, eg, lots of old disks hanging around, each too small to use alone, and so on.
Well, supossedly, raid5 is faster. I have a software raid1 (mirror) partition here, just for testing (I don't have any type of hardware or pseudo hardware raid), and it seems way slower than the rest of the non raid partitions. Let me see... (time dd_rescue /backup/boot2003.gho /dev/null) The raid partition reads at 20508kB/s Average (big file). Another partition reads at 24014kB/s. A write of one 204 Mb file times as: 7804kB/s (raid) 15995kB/s (normal) For these tests, not very scientific really, I used "time dd_rescue Mail.tar.gz ./delete" - because dd_rescue outputs some statitistics, although it is slow transferring. But it seems that write operations on a sofware raid 1 (mirror), two disks array, is half as slow in writing, and 25% slower when reading. The reading part I thought would be faster at least... In my case, both disks share the same cable (the other one is used for the cdwriter and the dvd - the CD forces the HDs to be terribly slow if placed on the same cable: no UDMA5 then). Probably the results would be different if I had the HD on separate cables, and perhaps even better as four disks as raid 5 on four different cables - but I don't have the hardware to test that, and I'm not buying it just for testing ;-) -- Cheers, Carlos E. R.