Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site tove.UUCP
Path: utzoo!watmath!clyde!bonnie!akgua!sdcsvax!sdcrdcf!hplabs!hao!seismo!umcp-cs!tove!mark
From: mark@tove.UUCP (Mark Weiser)
Newsgroups: net.lan
Subject: Re: Re: (DoD) InterNet Header Checksum
Message-ID: <13@tove.UUCP>
Date: Sat, 13-Oct-84 22:00:04 EDT
Article-I.D.: tove.13
Posted: Sat Oct 13 22:00:04 1984
Date-Received: Mon, 15-Oct-84 01:48:36 EDT
References: <624@ttds.UUCP> <4985@brl-tgr.ARPA> <1723@sun.uucp> <> <222@celerity.UUCP>
Organization: U of Maryland, Laboratory for Parallel Computation, C.P., MD
Lines: 16

> Considering how slowly the network code works (should have it's own dedicated
> cpu!) ...

We once thought that the network code took a significant amount of kernel time.
We have a vax 780 which serves as arpanet (via LHDH) gateway for about
10 heavily used machines on an ethernet.

4.2 unix kernel profiling says less than 1% of kernel cpu is spent handling
network gatwaying.  We had been considering putting the gatewaying on
a separate cpu (pdp-11/24) but can't see the point of it now.


-- 
Spoken: Mark Weiser 	ARPA:	mark@maryland
CSNet:	mark@umcp-cs 	UUCP:	{seismo,allegra}!umcp-cs!mark
U.S.: Computer Science Dept., University of Maryland, College Park, MD 20742