Path: newssvr20.news.prodigy.com!newsmst01.news.prodigy.com!prodigy.com!pd2nf1so.cg.shawcable.net!residential.shaw.ca!news-out.superfeed.net!propagator2-maxim!feed-maxim.newsfeeds.com!cyclone.bc.net!newsfeed.stanford.edu!postnews1.google.com!not-for-mail From: yong321@yahoo.com (Yong Huang) Newsgroups: comp.databases.oracle.server Subject: Re: VxFS block size with db block size =16k Date: 9 Sep 2003 19:11:05 -0700 Organization: http://groups.google.com/ Lines: 20 Message-ID: References: NNTP-Posting-Host: 24.123.141.249 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Trace: posting.google.com 1063159866 28582 127.0.0.1 (10 Sep 2003 02:11:06 GMT) X-Complaints-To: groups-abuse@google.com NNTP-Posting-Date: 10 Sep 2003 02:11:06 GMT Xref: newssvr20.news.prodigy.com comp.databases.oracle.server:242757 "Marco Vannini" wrote in message news:... > Hi, we 're managing a 8.1.7.4 VLDB (about 1TB) for DSS on SUN 3800, T3 > storage, Solaris 8, UFS with forcedirectio options for datafiles and raw > redo. Even if we're quite satisfied with the performance, we've purchased > Veritas Filesistem (not quick i/o), and considering to pass from UFS to > VxFS, maintaining the direct access (micache=direct convosync=direct). > Since the db_block_size=16k, is it better to make filesystems with 16k > block size or maintain the 8k of the actual UFS ? Hi, Marco, FYI, Bert Scalzo's test on Linux (published in Linux Journal) shows that 8k db_block_size gives faster data load than 4k, the default filesystem block size. My friend chao_ping's test on RH Linux AS using 9i tablespace multiple blocksize (un-published) shows that 16k has slight performance advantage over 4k in insert and significant advantage in full table scan, with db_block_size set to 8k. I would appreciate it if you could do a test and post the result here. Yong Huang