If I am not wrong Oracle randomly get free space regardless how many files you have, as far as the file is "writable". I think is documented, but I am not sure. ----- Original Message ----- From: "Wolfgang Breitling" <breitliw@xxxxxxxxxxxxx> To: <AmihayG@xxxxxxxxx> Cc: <oracle-l@xxxxxxxxxxxxx> Sent: Monday, August 01, 2005 2:43 AM Subject: Re: Q: oracle data files & performance > At 11:34 AM 7/31/2005, Amihay Gonen wrote: > >Hi , > >I've received a two interesting questions from a collage of mine. > > > >1) > >If two processes are writing to the same file , will they suffer > >from some conation on the file handler ? > >Do you think that spreading very busy tables (has a lot of inserts > >from different sessions) on different tablespaces (which will be > >translate to different data files ) will have better I/O performance > >over one single data file ? > > In Oracle no two processes write to the same datafile. Writing > changed blocks to the files is the job of the db writer (dbwr) and > thus a single process. Even if IO slaves are used, the dbwriter will > collect all blocks for the same file and give them to a single slave to write. > > > > >I'm talking only from performance perspective , not from backup & recovery ? > > > >2) > >If we have tablespaces with several data files . Will oracle > >allocate extents in a round-robin fashion between the files or will > >he fill one data file and then pass to the another data file ? > > I am not certain of the answer. I have seen indications of both - > well not exactly both, but uneven growth of the datafiles so not > exactly strict round-robin extent allocation but also not exactly > filling one file totally before using the other(s). > > Counterquestion - is Oracle a He or a She - or an It or a hermaphrodite > > > Regards > > Wolfgang Breitling > Centrex Consulting Corporation > http://www.centrexcc.com > > -- > //www.freelists.org/webpage/oracle-l > -- //www.freelists.org/webpage/oracle-l