From: "Howard J. Rogers" Newsgroups: comp.databases.oracle.server References: <9gqntt$p48$1@taliesin.netcom.net.uk> <3B30EF2D.BC7C6F49@attws.com> <9gsco2$e82$1@unbe.sarenet.es> <3b3f00ca@news.iprimus.com.au> <3B4756BE.75C016E6@home.com> Subject: Re: Urgently need help on: "Cannot Open Oracle DB when startup" Lines: 83 X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.50.4522.1200 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4522.1200 X-Original-NNTP-Posting-Host: 203.134.81.122 Message-ID: <3b47e38a@news.iprimus.com.au> X-Original-Trace: 8 Jul 2001 14:37:30 +1000, 203.134.81.122 Date: Sun, 8 Jul 2001 14:36:18 +1000 NNTP-Posting-Host: 203.134.64.67 X-Trace: news0.optus.net.au 994566981 203.134.64.67 (Sun, 08 Jul 2001 14:36:21 EST) NNTP-Posting-Date: Sun, 08 Jul 2001 14:36:21 EST Organization: CWO Customer - reports relating to abuse should be sent to abuse@cwo.net.au "Paul Drake" wrote in message news:3B4756BE.75C016E6@home.com... > "Howard J. Rogers" wrote: > > > So, my first thought is: does this guy have his redo logs mirrored within > > Oracle, and if not why not? Perhaps this experience will serve to underline > > the importance of three-way multiplexing of redo logs. > > > > Regards > > HJR > > Howard, > > Could you provide me with a link or a quick summary as to quantifying the > benefits of three-way-multiplexing of the online redo logs? I like 3-way multiplexing because the chances of LGWR stuffing up a write three times in a row must be smaller than it stuffing up twice in a row. More than three would also reduce the chances of that happening, obviously, but then you start running into potentially serious performance problems as, on most file systems, LGWR will be making *synchronous* writes... and hence there's a risk that the log buffer will fill up faster than LGWR can clear it. > > I can see the benefit of duplexing the online redo logs, and I know that Oracle > will use in a round-robin fashion the members of a group during the archival > process - but I cannot see the benefit outweighing the cost for a smallish > database. > Ah.. OK. So this is a discussion really about cost-benefit analyses, not about the merits of 3-way multiplexing per se? Fair enough. I guess that for many, the costs will outweigh the benefits (though having seen 2 out of 3 log members corrupted, but never having seen 3, and knowing the potential costs of corrupting a current redo log, I'd say the issue was still debateable). > for a small app running on a dedicated database that has at most 8 drives on 2 > I/O channels set up as 4 RAID 1 mirrored sets. > I'll probably be shot for saying so, but I'm not entirely convinced that the answer to that shouldn't be 'Oracle should not be running critical apps on low-end hardware such as you describe'. I guess we can debate this, too (though I speak as one who has just junked all the domestic Oracle databases I use for graphics stores, document repositories etc etc for a SQL Server implementation (1Gb RAM, 1GHz single cpu, 2 RAID-1 arrays) -much more suitable for my sort of home hardware, I think). > yes, I could see 3 way multiplexing on a 22 drive dream configuration. > Good! Then we agree... ;-) The short answer is that, yes, I tend to advocate dream configurations, without due regard for the pitfalls and shortfalls of the real world. Point taken. On the other hand, have you seen the latest Oracle licensing fees??! I reckon another RAID array and controller would be small beer for anyone who's forked that out recently! Regards HJR > thanks much, > > Paul Drake > >