On Fri, Oct 01, 1999 at 05:26:38PM +0100, Fred Backman wrote:
> I'm reading the web page about ucspi-tcp and see the last three lines which
> say:
>
> * inetd is unreliable under high loads. It cuts off service for 10 minutes
> if it receives ``too many'' connections in 1 minute.
In its default configuration, connect and disconnect from a service 50
times in one minute. inetd will now turn that service off for the next
10 minutes. This is how inetd is designed.
> * inetd does not provide effective resource management. It will happily use
> up all your memory if you are running a popular service.
Over the course of a long period of time, open up one connection to a
service per minute, and keep them all open. inetd will not limit the
number of connections and cause far too many (for example) POP3 servers
to be started. This is well known.
> * inetd has trouble with sudden bursts of activity. Its listen() backlog is
> typically only 5 or 10 and cannot be raised.
Try to simultaneously open 20 connections to a service and see how many
of them succeed, without the service being turned off. This is a little
more difficult, as those 20 really have to be nearly simultaneous to
trigger the problem.
> Where can I find further documentation and proof or witnesses of this
> actually being true?
I couldn't point you at specific attributions right now, but these are
well known problems, some of which are documented by the way inetd is
configured.
--
Bruce Guenter, QCC Communications Corp. EMail: [EMAIL PROTECTED]
Phone: (306)249-0220 WWW: http://www.qcc.sk.ca/~bguenter/
