X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9FD02D1BAF5 for ; Wed, 16 Jun 2004 17:18:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00519-05 for ; Wed, 16 Jun 2004 20:18:26 +0000 (GMT) Received: from mpls-qmqp-03.inet.qwest.net (mpls-qmqp-03.inet.qwest.net [63.231.195.114]) by svr1.postgresql.org (Postfix) with SMTP id AE378D1B1BE for ; Wed, 16 Jun 2004 17:18:19 -0300 (ADT) Received: (qmail 28564 invoked by uid 0); 16 Jun 2004 19:49:50 -0000 Received: from mpls-pop-11.inet.qwest.net (63.231.195.11) by mpls-qmqp-03.inet.qwest.net with QMQP; 16 Jun 2004 19:49:50 -0000 Received: from 63-227-127-37.dnvr.qwest.net (HELO ?10.0.0.2?) (63.227.127.37) by mpls-pop-11.inet.qwest.net with SMTP; 16 Jun 2004 20:18:20 -0000 Date: Wed, 16 Jun 2004 14:18:45 -0600 Message-Id: <1087417125.1582.47.camel@localhost.localdomain> From: "Scott Marlowe" To: "Bill" Cc: pgsql-performance@postgresql.org Subject: Re: Basic Postgresql Performance Question In-Reply-To: <1087413327.3207.9.camel@mathlab09.uchicago.edu> References: <1087413327.3207.9.camel@mathlab09.uchicago.edu> Content-Type: text/plain Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200406/164 X-Sequence-Number: 7222 On Wed, 2004-06-16 at 13:15, Bill wrote: > Hi, > I currently have a mysql server running with a database of around 800 > gb. The problem is that the server is old (500 MHz Pentium III with 512 > MB RAM) and I want to change this to a new server and convert the > existing database to Postgresql on Debian (I assume that Postgresql > offers better performance for complex read only queries on large > databases), Usually, but there are always queries that run faster or slower on a given database due to differences in architecture and design. For instance PostgreSQL tends to be slow when doing max/min aggs, but faster when doing things involving complex joins and unions. > though I was wondering if > > 1. It is possible to have some sort of load-balancing through buying > many computers without replication, i.e have one server which has the > databases and then other servers which has no database but just exists > to balance the memory and processor load? (I have heard this is possible > with C-JDBC)It is difficult to have enough space to replicate a 600 gb > database across all computers) That depends. Most databases are first I/O bound, then memory bound, then CPU bound, in that order. With an 800Gb database your main "cost" is gonna be moving data off of the platters and into memory, then having the memory to hold the working sets, then the CPU to mush it together. Now, if you're reading only tiny portions at a time, but doing lots of strange work on them, say weather forcasting, then you might be CPU bound. But without knowing what your usage patterns are like, we don't know whether or not running on multiple boxes would help. There are replication systems, but there's no such thing as a free lunch. > 2. It is advantageous to buy AMD 64 rather than the Pentium IV? Yes and no. If having more than 2 gigs of ram is important, 64 bit architecures run faster than 32 bit, where having over 2 gigs usually results in a slow down due to the memory switching they use.