Path: newssvr20.news.prodigy.com!newsmst01.news.prodigy.com!prodigy.com!logbridge.uoregon.edu!newsfeed.stueberl.de!ecngs!feeder.ecngs.de!zen.net.uk!peernews.manap.net!195.102.240.39.MISMATCH!newsr2.u-net.net!53ab2750!not-for-mail From: "Andy" Newsgroups: comp.databases.oracle.server References: Subject: Re: Single-disk database and I/O load balancing? Lines: 57 X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 6.00.2800.1106 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 Message-ID: Date: Fri, 10 Oct 2003 15:41:51 +0100 NNTP-Posting-Host: 213.2.27.2 X-Trace: newsr2.u-net.net 1065796755 213.2.27.2 (Fri, 10 Oct 2003 15:39:15 BST) NNTP-Posting-Date: Fri, 10 Oct 2003 15:39:15 BST Organization: (Posted via) VIA Net.Works UK Ltd. X-Received-Date: Fri, 10 Oct 2003 15:39:15 BST (newsr2.u-net.net) Xref: newssvr20.news.prodigy.com comp.databases.oracle.server:245016 Load balancing on one - sound theoretically perfectly balanced - it's perfectly balanced all on one disk :-)) Seriously using one disk for an oracle database is asking for trouble in most (all?) production environments. You've no safety margin - if that disk fails it's game over (at least whilst you find another disk and reinstall from a backup). Assuming disk wait events are the key events you're witnessing and the code is reasonably tuned and you're customers are complaining about performance then you need to explain that their system underperforms because they haven't got sufficient disk spindles AND that they are at risk of loosing data (assuming your one disk is not mirrored). It's not about total space but number of disk spindles and the amount of "work" taking place. I can imagine some of the regular contributors to this forum will have more to say when they recover from either the shock or the laughter. Typically I see a db of 100-400GB go on a disk subsystem with 8-16 disks in some configuration of RAID-5 & RAID-10 ( Usually with 18GB or 36GB disks ). If the customer is too "cost concious" to buy more a sufficient disk system then they probably shouldn't be using Oracle. If they don't need the resilience of a main stream database they probably shouldn't waste money buying it. If they do need it they need more disks. Andy "Waiting for the all the other opinions :-O" "Geomancer" wrote in message news:cf90fb89.0310100527.202279d8@posting.google.com... > With the Oracle10g Automatic Storage Management featre, I saw the > recommendation for "stripe and mirror everywhere". > > But what about smaller systems on a single, large disk? > > There are many Oracle databases that are less than 100 gigabytes, and > in my experience, a majority of Oracle systems are under 100 gig, and > most fit onto a single 74 gig or 144 gig disk drive. > > It does not make sense to stripe across a single disk, and I > experience disk euqueues, especially during large updates when undo > and data blocks are simultaneously updated. I also see read enqueues > when Oracle accesses index and table during the same query. > > I'm also curious if the newer disks like EMC and Hitachi have "fixed" > read-write heads? Our existing disk have moving read-write heads, and > this seek delay is over 80% of disk access latency. > > With the new 72 and 144 gig drives, it is hard to get my clients to > buy extra disks when the existing disk is large enough to hold their > entire database. These systems are always constrained by disk I/O. > > So, is the idea of disk load-balancing truly a fraud for non-RAIDED > disks?