[SunRescue] LAN Commmunication (was Sparc 10 cover plates)
GLeblanc at cu-portland.edu
Wed Dec 22 22:55:08 CST 1999
> -----Original Message-----
> From: woods at most.weird.com [mailto:woods at most.weird.com]
> Sent: Wednesday, December 22, 1999 8:20 PM
> To: rescue at sunhelp.org
> Subject: Re: [SunRescue] LAN Commmunication (was Sparc 10
> cover plates)
> Remember the good old days when everyone warned you against
> running such
> horrendous protocols like the old Unix rwhod? :-) If you
> didn't watch
> out it would make your network crawl to a halt! Oooooh! Scarrrry!
I'm afraid that's slightly before my time, I've only been into *nix for a
couple of years. :(
> On a slightly related note I found a "tech note" on NCD's site that
> purported to be about traffic engineering an Ethernet LAN for X11
> workstations and such. I didn't actually read it though..... :-)
I may have to go find that one...
> BTW, just exactly how many of you folks who are running big networks
> regularly get out the calculators (or spreadsheets! ;-) and work out
> exactly when and where you have to add in new switches and
> routers? :-)
I don't know if I'm running a "big network" or not, but here's my $0.02.
There are WAY too many variables that go into this to develop any algorithm
to figure it out. We basically have a star configuration, with a group of
switches at the center, and hubs at all other locations. The only time that
we'd have to add switches is when we add buildings, or perhaps when we add a
lot of computers in the existing buildings. If you get to design a lan from
the ground up, as was done here, expansion should be easy for quite some
time, unless the operation undergoes some MAJOR changes in network useage.
> I actually watch the traffic stats on all the ethernet ports on my
> switch and on each server and workstation, etc., (with SNMP), and even
> with all that SNMP traffic, my regularly busy e-mail server,
> DNS, X11 to
> a few workstations, including xclock with second hands on each, NTP,
> WWW, remote ping traffic watching my client sites, lots of telnet
> typing, the occasional NAS audio stream, NFS to my diskless
> and NFS between the development servers, etc., my little network is
> extremely boring! ;-) It is kind of handy though to debug problems
> because I can spot abnormal traffic patterns just by watching the
> blinking lights, and I can usually identify the culprits by comparing
> the traffic graphs for the switch ports and for each computer itself.
Sounds like you do some more detailed analysis than most "network experts"
that I know. I do practically no monitoring on our network, because there
are too many other things that have to get done (and also because I don't
REALLY know how to use a sniffer or SNMP).
More information about the rescue