Re: Scenario using SMB's RF option and different starting algorithm

mogul (Jeffrey Mogul) Fri, 01 December 1989 22:46 UTC

Received: by acetes.pa.dec.com (5.54.5/4.7.34) id AA09536; Fri, 1 Dec 89 14:46:04 PST
From: mogul (Jeffrey Mogul)
Message-Id: <8912012246.AA09536@acetes.pa.dec.com>
Date: 1 Dec 1989 1446-PST (Friday)
To: mtudwg
Cc:
Subject: Re: Scenario using SMB's RF option and different starting algorithm

Charlie Wickham writes:
	Do we know what the
	average lifetime of a VC is? Has anyone collected any stats on VC life?

Well, I just checked on several different machines here.  One is a
timesharing machine that does a lot of X-windows and telnet
connections; its connection lifetime
	 (packets sent+received)/connections_made
is about 1200 packets per connection.  One is decwrl.dec.com, which is
almost entirely a mail switch (and almost half of the mail connections
are non-local); it runs about 60 packets/connection (but decwrl was
rebooted only an hour ago, and so the average may be skewed by the
startup activities such as reloading the nameserver.)  Finally,
gatekeeper.dec.com, which is at least half used for non-local FTP and
Telnet, averages about 570 packets/connection.

Note that these are averages, and so they could be seriously skewed
by a few long-running connections.  For example, decwrl may be doing
some NNTP (netnews) transfers to internal hosts sending large batches
over a single connection.  It would be a good idea to try to measure
the distribution of connection lifetimes somehow.

-Jeff