sater@cs.vu.nl (Hans van Staveren) (02/19/88)
We have a collection of thin Ethernet segments connected by BICC Isolan Multiport repeaters. Since today we have a SUN 3/60 which seems to be able to do NFS and ND transfers with 6 packets back to back on the Ethernet. We have a suspicion that our repeaters can't cope with this. Looking at the various segments with etherfind, an ethernet monitoring program supplied by Sun, we saw packets on the originating segment, that didn't arrive on the destination segment. Does anybody have any experience with these repeaters, or others that suggest that we might be correct in our suspicion? Hans van Staveren
kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England)) (02/23/88)
In article <657@sater.cs.vu.nl> sater@cs.vu.nl (Hans van Staveren) writes: >We have a collection of thin Ethernet segments connected by >BICC Isolan Multiport repeaters. [...] > >We have a suspicion that our repeaters can't cope with this. >Looking at the various segments with etherfind, an ethernet monitoring >program supplied by Sun, we saw packets on the originating segment, >that didn't arrive on the destination segment. > A repeater has no buffering or processing to slow it up. The repeater can't be the source of your problems. However, a repeater also cannot isolate individual segments from faults and jabbering transceivers, etc. I suspect that you have excess collisions due to problems with the installation of your thin Ethernet. Etherfind/tcpdump will not find these problems. You need a LAN analyzer or a TDR to start. I recommend putting an analyzer on the troublesome segments to measure the collision rate. Once you know you have problems, TDR the segment and investigate the reflections. I recommend an analog TDR, not a digital meter. I suspect you have loose connectors or an unterminated segment or damaged cable. These problems don't necessarily evidence themselves until you strain the network, which your new server is probably doing. I bet you daisy-chained the thin cable through the walls and ceilings to save money, right? I've regretted ever doing that. Tracking down these kinds of problems is much harder. I hope you have keys for all those people on their winter vacations this week :-) Kent England, Boston University
ron@topaz.rutgers.edu (Ron Natalie) (02/23/88)
We find that the 3/60 as well as other suns will outdrive each other by sending ethernet packets so fast that the remote sun can not recieve. Using etherfind and traffic is no good as anything that can outdrive a SUN server is probably going to out drive your clients running etherfind as well. The program while very useful doesn't catch all the packets. -Ron