Path: news.easynews.com!easynews!news-out.cwix.com!newsfeed.cwix.com!news.maxwell.syr.edu!newsfeed1.uni2.dk!news.cybercity.dk!not-for-mail From: Svend Jensen Newsgroups: comp.databases.oracle.server Subject: Re: 8.1.7 unpredictable extreme slow exports on NT Date: Sun, 07 Jul 2002 18:10:02 +0200 Organization: Oracle Care Inc. Lines: 42 Message-ID: <3D2867DA.8090708@it.dk> References: <3d2723b9.10031294@news.hccnet.nl> NNTP-Posting-Host: port12.cvx1-kj.ppp.cybercity.dk Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Trace: news.cybercity.dk 1026058293 70398 217.157.72.13 (7 Jul 2002 16:11:33 GMT) X-Complaints-To: abuse@cybercity.dk NNTP-Posting-Date: Sun, 7 Jul 2002 16:11:33 +0000 (UTC) User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; CDonDemand; rv:0.9.4) Gecko/20011019 Netscape6/6.2 X-Accept-Language: en-us, da Xref: easynews comp.databases.oracle.server:153311 X-Received-Date: Sun, 07 Jul 2002 09:09:01 MST (news.easynews.com) 76434.1353@compuserve.com wrote: > The situation is like this: > > We have installed several Oracle databases on different location and > and hardware. IBM / Compaq > I read a lot about export problems performance problems on the net but > not mine. > What is the case: Normally the exports performed well dumps of 700 MB > in 4 minutes but sometimes the exports slow down and the same quantity > of data takes 1.5 hours. After a few days the export performes > suddenly normal again (4 minutes) > We do have this on every location and it is totally inpredictable when > it happens. It has nothing to do with daily operations because > everyday is the same and there are no end of months issues. > A shutdown and reboot does not make any diffrence. > Does anyone has an idea what the problem is? > > The problem is the unpredictable slow exports and the reason why? > Friendly regards, > > Ton > > How about delayed block clean out. When the db-writer gets behind, it doesn't cleanup, beautify and remove deleted rows (only removed from the row dictionary) and doesn't 'compress' the blocks before writing to disk. The block clean out is then left to who's ever reading the block from disk next time, in your case the export - reading all blocks from selected items to be exported. You can probe the clean out by analyze/compute statistics for some tables in the export set - time it. If its sometimes fast and sometimes slow - the block clean out is at stake here. Before export analyze again - if slow then delayed block clean out is the bad guy here. Then you can speed up the export by analyzing all items before exp is set to action. rgds /svend