Resent-From: postman (POSTGRES mailing list)
Resent-Message-Id: <9311012007.AA01250@postgres.Berkeley.EDU>
Sender: owner-postman@postgres.Berkeley.EDU
Date: Mon, 1 Nov 93 12:06:55 -0800
From: verghese@pepper.stanford.edu (Ben Verghese)
Message-Id: <9311012006.AA09799@pepper.Stanford.EDU>
To: postgres@postgres.berkeley.edu
Subject: concurrent access
Resent-To: postgres-dist
Resent-Date: Mon, 01 Nov 93 12:07:01 PST


I have ported Version 4.1 to IRIX. I was trying to run a TPC benchmark style
program on POSTGRES. There are relations BRANCH, TELLER, and ACCOUNT. Each
transaction updates an account, teller, and bank record. If I ran just one of
these applications, it ran fine. However, if I run two, the time for each
transaction just keeps increasing, with every transaction. Till it starts
timing out with deadlock detection. Each transaction is run between a BEGIN
and END. Also from running pagedoc on my BRANCH relation file, I find that it
is corrupted. BRANCH is the relation with the most concurrent access.

>From previous messages, I knew that transactions were quite slow. But my
observations seem to indicate that there is some sort of correctness problem
with the locking. Has this been observed by someone else ? I'm wondering if
this may be a result of a bug from my port.

Ben Verghese


********************************************************************************
Ben Verghese				Ph: 	(415) 723-4388 (office)
Computer Systems Lab				(415) 725-3732 (lab)
Electrical Engineering Dept.		Email:	verghese@pepper.stanford.edu
Stanford University
********************************************************************************