Received: from ext04.sra.co.jp (IDENT:root@ykh28DS23.kng.mesh.ad.jp [133.205.214.23]) by hub.org (8.9.3/8.9.3) with ESMTP id VAA20382 for ; Wed, 29 Sep 1999 21:24:02 -0400 (EDT) (envelope-from t-ishii@ext04.sra.co.jp) Received: from ext04.sra.co.jp (t-ishii@localhost [127.0.0.1]) by ext16.sra.co.jp (8.8.8/8.8.8) with ESMTP id KAA01268 for ; Thu, 30 Sep 1999 10:15:59 +0900 Message-Id: <199909300115.KAA01268@ext16.sra.co.jp> To: hackers@postgreSQL.org Subject: Announcement: pgbench-1.1 released From: Tatsuo Ishii Reply-To: t-ishii@sra.co.jp Date: Thu, 30 Sep 1999 10:15:59 +0900 Sender: t-ishii@ext04.sra.co.jp Hi, I have made a small tool called "pgbench," that may be useful for some stress testings and performance measurements for PostgreSQL. pgbench can be obtained from: ftp://ftp.sra.co.jp/pub/cmd/postgres/pgbench/ Pgbench runs small transactions similar to TPC-B concurrently and reports the number of transactions actually done per second (tps). Pgbench uses the asynchronous functions of libpq to simulate concurrent clients. Example outputs from pgbench are shown below: number of clients: 4 number of transactions per client: 100 number of processed transactions: 400/400 tps = 19.875015(including connections establishing) tps = 20.098827(excluding connections establishing) (above result was reported on my PowerBook with 603e CPU(180MHz), 80MB mem running PostgreSQL 6.5.2 with -F option, and Linux 2.2.1 kernel) Pgbench does not require any special libraries other than libpq. It comes with a configure script and should be very easy to build. *CAUTION* pgbench will blow away tables named accounts, branches, history and tellers. It is best to create a new database for pgbench before running it. BTW, the greatest tps I have ever seen was around 260 on a Linux box running RedHat 6.0, having 2 Pentiumn III 600MHz CPUs and 512MB mem. Enjoy, --- Tatsuo Ishii