Path: news.easynews.com!easynews!news-out.cwix.com!newsfeed.cwix.com!newsfeed.mathworks.com!wn3feed!wn2feed!worldnet.att.net!204.127.198.203!attbi_feed3!attbi.com!sccrnsc02.POSTED!not-for-mail Message-ID: <3D291A93.E6587A0F@ureach.com> From: Karen X-Mailer: Mozilla 4.72 [en] (Win98; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.databases.oracle.server Subject: Re: Sparse TEMP files References: <3D27E270.D86488D9@ureach.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 28 NNTP-Posting-Host: 12.234.110.174 X-Complaints-To: abuse@attbi.com X-Trace: sccrnsc02 1026104025 12.234.110.174 (Mon, 08 Jul 2002 04:53:45 GMT) NNTP-Posting-Date: Mon, 08 Jul 2002 04:53:45 GMT Organization: AT&T Broadband Date: Mon, 08 Jul 2002 04:53:45 GMT Xref: easynews comp.databases.oracle.server:153351 X-Received-Date: Sun, 07 Jul 2002 21:51:08 MST (news.easynews.com) Actually I think you are right and I was wrong. It should be actually FASTER for temp files. For temp files, the OS blocks do not exist, and OS will not have to read them in first! I do not think that modifying inode pointers or formating the blocks by Oracle will be any significant overhead, as both are in memory operations. However, Oracle's implementation may turn our speculations upside down. For my organization, though, the monitoring considerations prevail and I would rather do what I suggested. Yong Huang wrote: > Karen wrote in message news:<3D27E270.D86488D9@ureach.com>... > > b) the performance > > will be slower for the first queries that use the tablespace. > > I'd like to know why the first use of a sparse tempfile would be > slower than if the file were actually filled with data. If it *is* > indeed true, it's probably not because the "holes" have to be written > with real data; it may be because the inode pointers have to point > somewhere instead of nowhere? Do you have a benchmark or a theoretical > proof? > > Yong Huang