From owner-freebsd-hackers  Mon Dec 22 09:32:09 1997
Return-Path: <owner-freebsd-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id JAA08546
          for hackers-outgoing; Mon, 22 Dec 1997 09:32:09 -0800 (PST)
          (envelope-from owner-freebsd-hackers)
Received: from acroal.com (firewall0.acroal.com [209.24.61.154])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id JAA08486
          for <hackers@FreeBSD.ORG>; Mon, 22 Dec 1997 09:31:58 -0800 (PST)
          (envelope-from jamil@acroal.com)
Received: from localhost (jamil@localhost)
	by acroal.com (8.8.8/8.8.7) with SMTP id JAA03319;
	Mon, 22 Dec 1997 09:31:41 -0800 (PST)
	(envelope-from jamil@acroal.com)
Date: Mon, 22 Dec 1997 09:31:41 -0800 (PST)
From: "J. Weatherbee - Senior Systems Architect" <jamil@acroal.com>
To: Mike Smith <mike@smith.net.au>
cc: hackers@FreeBSD.ORG
Subject: Re: Am I off my rocker? (/dev/tick device) 
In-Reply-To: <199712220950.UAA00717@word.smith.net.au>
Message-ID: <Pine.BSF.3.96.971222092648.3294A-100000@acroal.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

> > 2) On the vectorized i/o, think about processes multiplexing
> > with say 1000 - 2000 descriptors.
> 
> So?  Async I/O is *more* efficient, not less.

It is not an issue of asynchronous versus synchronous i/o.  As in both
cases you are still going to need to use a select() or poll() call (in the
signal handler).   Rather, it is an issue of going through the call gate x
number of times, in the case of a large number of descriptors.  And aside
from that (mikey) signals are expensive to catch [Leffler et al. 1989]
(Check Richard Stevens, Unix Network Programming, 1990, p. 328.