raid5 again
Floyd asked whether it's sensible, on a new IBM FastT-700 SAN, to put older read-only chunks on RAID 5 and keep active/growing chunks on RAID 10. Art Kagel argued no: RAID 10 beats RAID 5 on reads as well as writes (parallel reads from mirrored pairs), and more importantly RAID 5 risks data loss from partial media failure — especially for rarely rewritten historical data — pointing to baarf.com. Others noted vendors push RAID 5 and one poster reported no I/O problems with RAID 5 in practice; another simply bought raw disk and carved his own RAID. No definitive resolution is recorded — the thread is an opinion debate, with the consensus advice being to avoid RAID 5.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Since we are upgrading to an IBM FAST-700, supersan, and we are going to reorganize where our chunks are getting accessed from, we've been thinking. Yes, the general rule is that raid5 is bad, and Art, I know you hate it.... But. Raid5 is good for reading, correct ? A lot of our data is old and doesn't change. It get's read a lot but not written to. So why not put our full chunks on raid 5 and our newest chunks and growth on raid 10. Any thoughts ? Thanks, floyd ======================== -<<Floyd Wellershaus>>- Database Administrator Unix Administrator email: fwellers@yahoo.com Home: 703-430-0805 Cell: 703-477-6045 ========================
RAID10 will outperform RAID5 in read tests also, not just writes. In heavy random access reading RAID10 can read 2 independent sectors from each drive pair in parallel while RAID5 can read one sector from each drive. An intelligent array can reach close to 2x the read speed on sequential reads from RAID10 versus RAID5. My biggest problem is that one's data is not safe on the RAID5 array. Historical data will be around a long time, presumably, without being rewritten or reorganized, and so is even more suceptible to damage by partial media failure (the bane of RAID5) than active data simply because the magnetic flux changes are never refreshed. I don't care what the circumstances are, if you like your data, your employer, and being employed then there's only one mantra to chant: NO RAID5!! Art S. Kagel NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! ----- Original Message ----- From: Floyd Welle.... <fwellers@yahoo.com> At: 3/16 13:25 > Since we are upgrading to an IBM FAST-700, supersan, and we are going to > reorganize where our chunks are getting accessed from, we've been thinking. > Yes, the general rule is that raid5 is bad, and Art, I know you hate it.... But. > Raid5 is good for reading, correct ? > A lot of our data is old and doesn't change. It get's read a lot but not written > to. So why not put our full chunks on raid 5 and our newest chunks and growth on > raid 10. > Any thoughts ? > > Thanks, > > floyd >
First off, I AGREE! But as an aside, I have been meeting with IBM, EMC and a couple other storage companies and they Keep saying ALL our DB customers use RAID 5, no one uses 10 anymore, there is no need. Our system administrators quit inviting me to meetings....tired of hearing me rant and foam at the mouth. >>sigh<< S -----Original Message----- From: forum.subscriber@iiug.org [mailto:forum.subscriber@iiug.org] On Behalf Of ART KAGEL, .... Sent: Wednesday, March 16, 2005 1:44 PM To: ids@iiug.org Subject: Re: raid5 again [4521] RAID10 will outperform RAID5 in read tests also, not just writes. In heavy random access reading RAID10 can read 2 independent sectors from each drive pair in parallel while RAID5 can read one sector from each drive. An intelligent array can reach close to 2x the read speed on sequential reads from RAID10 versus RAID5. My biggest problem is that one's data is not safe on the RAID5 array. Historical data will be around a long time, presumably, without being rewritten or reorganized, and so is even more suceptible to damage by partial media failure (the bane of RAID5) than active data simply because the magnetic flux changes are never refreshed. I don't care what the circumstances are, if you like your data, your employer, and being employed then there's only one mantra to chant: NO RAID5!! Art S. Kagel NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! ----- Original Message ----- From: Floyd Welle.... <fwellers@yahoo.com> At: 3/16 13:25 > Since we are upgrading to an IBM FAST-700, supersan, and we are going > to reorganize where our chunks are getting accessed from, we've been thinking. > Yes, the general rule is that raid5 is bad, and Art, I know you hate it.... But. > Raid5 is good for reading, correct ? > A lot of our data is old and doesn't change. It get's read a lot but > not written > to. So why not put our full chunks on raid 5 and our newest chunks and > growth on > raid 10. > Any thoughts ? > > Thanks, > > floyd >
Floyd Welle.... said: > Since we are upgrading to an IBM FAST-700, supersan, and we are going to > reorganize where our chunks are getting accessed from, we've been > thinking. > Yes, the general rule is that raid5 is bad, and Art, I know you hate > it.... But. > Raid5 is good for reading, correct ? > A lot of our data is old and doesn't change. It get's read a lot but not > written to. So why not put our full chunks on raid 5 and our newest chunks > and growth on raid 10. > Any thoughts ? Yeah, but they'd only get me in shit with June again. -- Bye now, Obnoxio "C'est pas parce qu'on n'a rien à dire qu'il faut fermer sa gueule" - Coluche "I'm trying to see things your way, but I can't get my head up my ass" - JCH "Ogni uomo mi guarda come se fossi una testa di cazzo" - Marco Travel broadens a person. You look as if you have been all over the world. I went to the airport to check in and they asked what I did because I looked like a terrorist. I said I was a comedian. They said, "Say something funny then." I told them I had just graduated from flying school. -- Ahmed Ahmed http://members.rogers.com/2remedy/images/smilies/banners/RTFM.gif
There was a chain email regarding RAID 5, a few weeks back. I see a lot of passion for RAID 10. Sys Admin / Management don't want the DBs to use more disks - Right ? In my experience RAID 5 isn't bad. 1. We had Teradata on RAID 5 NCR machine. It was OK. 2. Now we have 2 instances (OLTP and DW) of IDS 9.4 on RAID 5 and we have not IO waits. Disks and Controllers (along with cache) have come a long way. RAID 5 performs as better as the RAID 10 now a days. (OLTP instance is a VERY high transactional system. DW is a medium size Instance > 300G.) 3. Multiple Oracle Instances (DBs) run on RAID 5. No performance degradation because of IO. I don't mean to offend anyone, it is just my view. Thought, I could share it with you guys. Thank you, Kannan Thirugnanam -----Original Message----- From: forum.subscriber@iiug.org [mailto:forum.subscriber@iiug.org] On Behalf Of Gentsch, Sam Sent: Wednesday, March 16, 2005 3:22 PM To: ids@iiug.org Subject: RE: raid5 again [4523] First off, I AGREE! But as an aside, I have been meeting with IBM, EMC and a couple other storage companies and they Keep saying ALL our DB customers use RAID 5, no one uses 10 anymore, there is no need. Our system administrators quit inviting me to meetings....tired of hearing me rant and foam at the mouth. >>sigh<< S -----Original Message----- From: forum.subscriber@iiug.org [mailto:forum.subscriber@iiug.org] On Behalf Of ART KAGEL, .... Sent: Wednesday, March 16, 2005 1:44 PM To: ids@iiug.org Subject: Re: raid5 again [4521] RAID10 will outperform RAID5 in read tests also, not just writes. In heavy random access reading RAID10 can read 2 independent sectors from each drive pair in parallel while RAID5 can read one sector from each drive. An intelligent array can reach close to 2x the read speed on sequential reads from RAID10 versus RAID5. My biggest problem is that one's data is not safe on the RAID5 array. Historical data will be around a long time, presumably, without being rewritten or reorganized, and so is even more suceptible to damage by partial media failure (the bane of RAID5) than active data simply because the magnetic flux changes are never refreshed. I don't care what the circumstances are, if you like your data, your employer, and being employed then there's only one mantra to chant: NO RAID5!! Art S. Kagel NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! ----- Original Message ----- From: Floyd Welle.... <fwellers@yahoo.com> At: 3/16 13:25 > Since we are upgrading to an IBM FAST-700, supersan, and we are going > to reorganize where our chunks are getting accessed from, we've been thinking. > Yes, the general rule is that raid5 is bad, and Art, I know you hate it.... But. > Raid5 is good for reading, correct ? > A lot of our data is old and doesn't change. It get's read a lot but > not written > to. So why not put our full chunks on raid 5 and our newest chunks and > growth on > raid 10. > Any thoughts ? > > Thanks, > > floyd >
Speed is the least of the problems with RAID5. Check out my posting and those of others, including some Oracle insiders, on www.baarf.com Art S. Kagel ----- Original Message ----- From: Thirugnanam.... <Kannan.Thirugnanam@JTax.com> At: 3/16 17:02 > There was a chain email regarding RAID 5, a few weeks back. > I see a lot of passion for RAID 10. > Sys Admin / Management don't want the DBs to use more disks - Right ? > > In my experience RAID 5 isn't bad. > 1. We had Teradata on RAID 5 NCR machine. It was OK. > 2. Now we have 2 instances (OLTP and DW) of IDS 9.4 on RAID 5 and we > have not IO waits. Disks and Controllers (along with cache) have come a > long way. RAID 5 performs as better as the RAID 10 now a days. (OLTP > instance is a VERY high transactional system. DW is a medium size > Instance > 300G.) > 3. Multiple Oracle Instances (DBs) run on RAID 5. No performance > degradation because of IO. > > I don't mean to offend anyone, it is just my view. Thought, I could > share it with you guys. > > Thank you, > Kannan Thirugnanam > > -----Original Message----- > From: forum.subscriber@iiug.org [mailto:forum.subscriber@iiug.org] On > Behalf Of Gentsch, Sam > Sent: Wednesday, March 16, 2005 3:22 PM > To: ids@iiug.org > Subject: RE: raid5 again [4523] > > First off, I AGREE! > > But as an aside, I have been meeting with IBM, EMC and a couple other > storage companies and they > Keep saying ALL our DB customers use RAID 5, no one uses 10 anymore, > there is no need. > > Our system administrators quit inviting me to meetings....tired of > hearing me rant and foam at the mouth. > > >>sigh<< > > > S > > > -----Original Message----- > From: forum.subscriber@iiug.org [mailto:forum.subscriber@iiug.org] On > Behalf Of ART KAGEL, .... > Sent: Wednesday, March 16, 2005 1:44 PM > To: ids@iiug.org > Subject: Re: raid5 again [4521] > > > RAID10 will outperform RAID5 in read tests also, not just writes. In > heavy random access reading RAID10 can read 2 independent sectors from > each drive pair in parallel while RAID5 can read one sector from each > drive. An intelligent array can reach close to 2x the read speed on > sequential reads from RAID10 versus RAID5. My biggest problem is that > one's data is not safe on the RAID5 array. Historical data will be > around a long time, presumably, without being rewritten or reorganized, > and so is even more suceptible to damage by partial media failure (the > bane of RAID5) than active data simply because the magnetic flux changes > are never refreshed. > > I don't care what the circumstances are, if you like your data, your > employer, and being employed then there's only one mantra to chant: NO > RAID5!! > > Art S. Kagel > > NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO > RAID5!! > > > ----- Original Message ----- > From: Floyd Welle.... <fwellers@yahoo.com> > At: 3/16 13:25 > > > Since we are upgrading to an IBM FAST-700, supersan, and we are going > > > to reorganize where our chunks are getting accessed from, we've been > thinking. > > Yes, the general rule is that raid5 is bad, and Art, I know you hate > it.... > But. > > Raid5 is good for reading, correct ? > > A lot of our data is old and doesn't change. It get's read a lot but > > not > written > > to. So why not put our full chunks on raid 5 and our newest chunks and > > > growth > on > > raid 10. > > Any thoughts ? > > > > Thanks, > > > > floyd > > > > > > > > >
We are moving to P-Series later this year with FastT storage. I simply told IBM how much basic disk I wanted and said I would carve it up my- self in whatever RAID I thought appropriate. No problems. I even managed small(ish ,36 Gb) disks to give me a decent number of spindles. I do have the advantage of being Sys Admin and DB Admin and although I might be tettering on the edge, I'm not schizophrenic yet. Keith -> -----Original Message----- -> From: Gentsch, Sam [mailto:sgentsch@intercall.com] -> Sent: Wednesday, March 16, 2005 8:22 PM -> To: ids@iiug.org -> Subject: RE: raid5 again [4523] -> -> -> First off, I AGREE! -> -> But as an aside, I have been meeting with IBM, EMC and a couple other -> storage companies and they -> Keep saying ALL our DB customers use RAID 5, no one uses 10 anymore, -> there is no need. -> -> Our system administrators quit inviting me to meetings....tired of -> hearing me rant and foam at the mouth. -> -> >>sigh<< -> -> -> S -> -> -> -----Original Message----- -> From: forum.subscriber@iiug.org [mailto:forum.subscriber@iiug.org] On -> Behalf Of ART KAGEL, .... -> Sent: Wednesday, March 16, 2005 1:44 PM -> To: ids@iiug.org -> Subject: Re: raid5 again [4521] -> -> -> RAID10 will outperform RAID5 in read tests also, not just writes. In -> heavy random access reading RAID10 can read 2 independent -> sectors from -> each drive pair in parallel while RAID5 can read one sector -> from each -> drive. An intelligent array can reach close to 2x the read speed on -> sequential reads from RAID10 versus RAID5. My biggest -> problem is that -> one's data is not safe on the RAID5 array. Historical data will be -> around a long time, presumably, without being rewritten or -> reorganized, -> and so is even more suceptible to damage by partial media -> failure (the -> bane of RAID5) than active data simply because the magnetic -> flux changes -> are never refreshed. -> -> I don't care what the circumstances are, if you like your data, your -> employer, and being employed then there's only one mantra to -> chant: NO -> RAID5!! -> -> Art S. Kagel -> -> NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO RAID5!! NO -> RAID5!! -> -> -> ----- Original Message ----- -> From: Floyd Welle.... <fwellers@yahoo.com> -> At: 3/16 13:25 -> -> > Since we are upgrading to an IBM FAST-700, supersan, and -> we are going -> -> > to reorganize where our chunks are getting accessed from, -> we've been -> thinking. -> > Yes, the general rule is that raid5 is bad, and Art, I -> know you hate -> it.... -> But. -> > Raid5 is good for reading, correct ? -> > A lot of our data is old and doesn't change. It get's read -> a lot but -> > not -> written -> > to. So why not put our full chunks on raid 5 and our -> newest chunks and -> -> > growth -> on -> > raid 10. -> > Any thoughts ? -> > -> > Thanks, -> > -> > floyd -> > -> -> -> -> -> -> -> -> -> ******************************************************************************** ** This message is sent in strict confidence for the addressee only. It may contain legally privileged information. The contents are not to be disclosed to anyone other than the addressee. Unauthorised recipients are requested to preserve this confidentiality and to advise the sender immediately of any error in transmission. This footnote also confirms that this email message has been swept for the presence of computer viruses, however we cannot guarantee that this message is free from such problems. ******************************************************************************** **