From owner-freebsd-current Mon Mar 2 23:56:22 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id XAA22961 for freebsd-current-outgoing; Mon, 2 Mar 1998 23:56:22 -0800 (PST) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from smtp03.primenet.com (smtp03.primenet.com [206.165.6.133]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id XAA22898; Mon, 2 Mar 1998 23:56:18 -0800 (PST) (envelope-from tlambert@usr05.primenet.com) Received: (from daemon@localhost) by smtp03.primenet.com (8.8.8/8.8.8) id AAA18080; Tue, 3 Mar 1998 00:56:17 -0700 (MST) Received: from usr05.primenet.com(206.165.6.205) via SMTP by smtp03.primenet.com, id smtpd018074; Tue Mar 3 00:56:12 1998 Received: (from tlambert@localhost) by usr05.primenet.com (8.8.5/8.8.5) id AAA17826; Tue, 3 Mar 1998 00:56:12 -0700 (MST) From: Terry Lambert Message-Id: <199803030756.AAA17826@usr05.primenet.com> Subject: Re: 3.0-RELEASE? To: dyson@FreeBSD.ORG Date: Tue, 3 Mar 1998 07:56:12 +0000 (GMT) Cc: tom@uniserve.com, current@FreeBSD.ORG In-Reply-To: <199803030601.BAA00325@dyson.iquest.net> from "John S. Dyson" at Mar 3, 98 01:01:13 am X-Mailer: ELM [version 2.4 PL25] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > > It is kinda of interesting that Solidtech ported their Solid server to > > FreeBSD 2.2, but performance is so-so because of lack of threading. > > That does not surprise me. Oracle is the same way (whoops!!!), because > the current FreeBSD port doesn't use AIO. (The port was done a long time > ago.) It is mostly used in the NC server, and doesn't really need high > perf. > > Actually, AIO should help databases alot. Yes. I was going to point out that user space threading actually works up to Draft 4 requirements now, and there is very little difference between thre and draft 10. The main difference in using AIO or kernel threads is that you can interleave read operations (write operations are interleaved anyway). There's a bit of argument possible here; for example, when you make a read with O_NONBLOCK, and it returns EWOULDBLOCK, it's probably a good thing to trigger a read-ahead. This would allow the current user space implementation to interleave I/O as effectively as a kernel threads implementation, and with an additional NULL system call overhead relative to each read call. An AIO may or may not be able to cluster waits in the average case, so this may be a wash. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message