goetz@reed.UUCP (Norman Goetz) (09/16/90)
We have a LocalTalk network of 150 nodes (users + servers) running on 4 Farallon StarControllers. The wiring meets the Farallon specs, the users do not see any slowness due to heavy traffic, and the monitored network traffic reaches 100% for only a few seconds at a time on few occasions. QuickMail, AppleShare, and two hard disc backup programs running at night are the traffic. Users are booted up during the day but many of them only occasionally use their Macs. An fx was recently added, and is running slowly, but we have ordered the new ROMs from Farallon that are supposed to help. We also see spontaneous restarting of our FastPath from zero-length packets but this is a ROM bug which Shiva is supposed to cure in the future. We also see some of the Mac Pluses finding a node number in the upper address range, 128-254, presumably because the lower range is relatively full. This behavior was not permitted in the original "Inside AppleTalk" but was changed in the ROMs after 1987, and is now permitted according to Apple. Seeing 150 nodes, most people would install a bridge. We own one and are considering it, but the users are happy and none of the problems of the previous paragraph seem to be due to node number crowding. My question: is there some subtle problem we have or are about to have due to 150 nodes, to having user nodes in the upper address range, or to running this situation through StarControllers? Norman Goetz, Network Technician, Reed College goetz@reed.bitnet -- Norman Goetz Network Technician goetz@reed.bitnet audio: (503) 771-1112 X 646 "This has been a test life. This was only a test. If this had been an actual life you would have been given instructions on where to go and what to do."
Beattie@SYSTEM-M.PHX.BULL.COM ("Art Beattie {AC602 862-4783}") (09/19/90)
Norman Goetz (goetz@reed.bitnet) writes:
My question: is there some subtle problem we have or are about to
have due to 150 nodes, to having user nodes in the upper address
range, or to running this situation through StarControllers?
You seem to have a similar user group that we have. Whenever any one of
our zones approaches having 100 active nodes, we start thinking of
partitioning the one zone into 2. The list of printers in the Chooser
window starts to really jump and users get a bit concerned. Sometimes
their printer doesn't show up in the list for a couple of minutes.
I am assuming that all 150 users are on one LocalTalk zone. Not sure
what is on the other side of the FastPath. At this point I would look
into getting another FastPath or Gatorbox or MultiGate and partition the
zone into 2 75 node zones.
Realize that there is *ALWAYS* traffic on the LocalTalk wiring. The
various protocols are banging away on the zone maintaining tables and
making sure other zones are still there, etc. A LocalTalk network is
very suspectible to congestion problems because of its speed. It
operates very similarly to ethernet (collision recovery, etc) but due to
is speed, these mechanisms are slow. It only takes 2 or 3 of users
doing file transfers to bring a LocalTalk zone to its knees. LANRanger
can show you the stats on collision traffic. It doesn't take much.
Good luck.