J1C6844@TAMSIGMA.BITNET.UUCP (04/18/87)
At Texas A&M University our Ethernet is fairly responsive. With data from
packages like Ethernim, and hardware such as LANalyzers and HP's LAN analyzer,
we noted that we occasionally use 5% of the bandwidth, and rarely use more than
10%. Of course as someone has already mentioned it is not possible to have 100%
utilization of the 10Mbit theoretical bandwidth ( my guess would be 35% - 50% ).
Every so often you can notice a slight pause during a particular heavy moment of
traffic. Currently we have connected to our Ethernet:
60 DECnet nodes (from DECpros to an 8800)
50 TCP/IP nodes (from Apollos to Amdahls)
35 Terminal servers ( DECserver 100s, 200s, DECSAs, Bridge
Communication Servers)
Also, we have many PC's, a DECSA router, a SNA Gateway, and a Local Area
Cluster on one segment. The network comprises of approximately 25 coax segments
linked with fiber optics through remote repeaters and a few local repeaters.
There are 8 or 9 bridges that separate the network at strategic points. The
majority of the traffic is due to LAT protocol (80%-90%) as approximately 35 of
the nodes offer LAT services. There are upwards of 20 reverse LAT print queues
used throughout the network and a Printserver 40 was recently installed. Even
with all of this the statistics above hold true for the most part. The number
of attached items changes daily so the numbers above are not exact. There are
around 50,000 students, faculty and staff at the university and some of the
machines have upwards of 6000 accounts on them. There is probably at least 1500
terminals at various places on campus to access the network. This network was
well managed from the start by a manager of one of the VAXes and continues
to function very well. From the above facts it would appear that the Ethernet
is capable of handling a great deal of traffic, but it doesn't do it all by
itself, the Ethernet must be managed. I hope this will shed some light on
the capabilities of an Ethernet network.