Summary of my original question:
----------------------------------------------------------------------
My LAN experts today told me that my team is pushing our LAN load over
the maximum through-put point such that we have enough collisions to
significantly reduce actual data movement. I have run far more traffic
on ethernets with non-Sun hardware and software, so I am rather
surprised by this LAN usage analysis. Before we start taking any
actions, I wanted to ask all you managers familiar with running
SPARCstations on ethernet LANs the following questions:
---------------------------------------------------------------------
While I am still getting some replies (keep them coming if you think
you can add to what I say in this summary), I think the trend is
pretty clear.
1. I got a lot of "me too"s.
2. Standard ethernet is not designed for the sustained data rates that
groups of SPARCstation 20s and 10s can easily reach. Most striking was
a statement that if I start a network copy from one SS20 to another,
that operation alone could produce a sustained traffic level of 50% of
an ethernet. Given the nature of ethernet, an operation between a
second pair of machines, even if small, have a high likelyhood of
producing collisions. Add a few more machines into the mix and we are
quickly on the down-side of the ethernet through-put curve.
3. Look for bad a ethernet cable, a bad connector, a bad hub
component, or a bad terminator.
4. Tools available: netstat -i, performance meter, nfsstat, iostat,
etherfind, and a number of votes for Etherman... (if someone could
give me an ftp pointer for etherman, I would be grateful)
5. The only solution is new hardware, of which a large variety of
solutions were described.
Marc S. Gibian
Telos Consulting Services phone: (617) 377-6350
PRISM/TFS email: gibian@stars1.hanscom.af.mil
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:18 CDT