Re: SSD --- The reliability of

Penang schrieb: > Please pardon me if this topic has been discussed before, but as far > as I know about the Flash memory, they have a relatively low limit on > re-write, as compared to hard disk > > In situation where database use in concern, read and write operation > takes place very often > > I know that there are pros and cons in this issue, but basically my > question is this --- How reliable is SSD? Hi Penang, sorry for joining in so late, but I did not have access to usenet the last 4 days. You will find a l o t of useful information at www.storagesearch.com To make it short: A flash disk is not a real SSD. Real SSDs do like 16K IOPS per disk and they are very reliable, very expensive and on the market for like 20+ years. There are 2 types of flash disks: SLC (single layer) - smaller (i.e.upto 64GB), more expensive and faster. MLC (multi layer) - those are bigger (upto 512GB from Toshiba) usually 256 GB now and in the EUR 600 price range for 256 GB. So if a flash disk 'wears out' first thing to happen is, that is gets slower, much slower, that is, and in a short period of time (exactly this is Intel's problem atm). After a long period they finally are so slow, that I think they might never die from wearing out ..... I am not that far in testing, though. What I do in week 9 atm is trying to do harm to the ones I testing. I give them 2800 IOPS writing only and 2 KB blocks only. I have to types: internally having 4 KB blocks or 8 KB blocks. So I write 86400 x 2800 blocks per day for 64 days now. None is slower that at the beginning - so far. If you can throw in a lot of money, have a look at the real speed: Texas Memory Systems www.ramsan.com Or on a much cheaper scale: the 12 slot boxes from www.infortrend.com (in heavy use at CERN, Switzerland). And n e v e r e v e r go RAID5 using a flash disk! Writing the parity there kills a third of the lifetime. www.baarf.com has all on RAID5 and why never it is a good idea. dic_k -- Richard Kofler SOLID STATE EDV Dienstleistungen GmbH Vienna/Austria/Europe

On Fri, 14 Aug 2009 22:50:17 +0200, Richard Kofler <richard.kofler@chello.at> wrote: >And n e v e r e v e r go RAID5 using a flash disk! >Writing the parity there kills a third of the lifetime. >www.baarf.com has all on RAID5 and why never it is a good >idea. In some cases a third of the lifetime would be a trivial amount to lose. The big mistake many seem to make when considering flash limited write-cycles is they are trying, deliberately trying, to fill the whole drive. Instead of leaving a few GB of free space, if a drive is selected that has 4X as much free space, you quarter the write cycles. That could be a substantial difference, supposing it might last 16 years instead of 4. I don't mean to suggest flash is a universal replacement for other solid state storage but for most uses it is more reliable.