Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!ucbvax!NAZGUL.PHYSICS.MCGILL.CA!loki
From: loki@NAZGUL.PHYSICS.MCGILL.CA (Loki Jorgenson Rm421)
Newsgroups: comp.sys.sgi
Subject: network collision poll
Message-ID: <9104210057.AA01148@nazgul.physics.mcgill.ca>
Date: 21 Apr 91 00:57:42 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 33


	I suspect that I may have some sort of problem in my network.  This
feeling is based on two observations:

	1)  Sometimes, IRI acting as NFS disk servers are adversely affected by 
	I/O intensive programs running on other IRI writing to disk; the
	nfsd consume most of the cpu time and write rates fall to kbyte/minute
	levels. 

	2) Collision rates on all IRI are between 2% and 10% of total in-bound
	and out-bound packets (compared to 0.2% on all SUN3s on the same net).
	These rates are according to "netstat -i".

	I would like suggestions as to what I should look for BESIDES cables
of incorrect length and bad BNC connections.  Also, if you have a comparable
network arrangement (9 diskless SUN 3/50s served by a 3/180, a diskful 3/60,
3 4D/25, a 4D/20, a 4D/340 and 2 NeXT, all on thinnet; 3/180 is the gateway),
please send me the results of a "netstat -i" for your IRI.  Prehaps 5%
collisions is normal.

	Also, is my understanding that Ethernet cables should be
multiples of 50cm in length correct?  Is there a minimum length?
What is your opinion of a maximum length?  I consider these sorts
of numbers to be black magic and subject to context but I would
like your experiences.

	I shall summarize if I get enough useful info.
Thanks,
                             __          __
Loki Jorgenson              / /          \ \  node:  loki@Physics.McGill.CA
Grad, Systems Manager      / //////  \\\\\\ \ BITNET: PY29@MCGILLA
Physics, McGill University \ \\\\\\  ////// / fax:   (514) 398-8434
Montreal Quebec CANADA      \_\          /_/  phone: (514) 398-7027