cyrus@hi.unm.edu (Tait Cyrus) (04/02/88)
I am looking for an equation that roughly computes the theoretical
load on a network given only the following information:
1) time over which the load is to be computed,
2) the number of packets seen in that time,
3) the number of bytes seen in that time.
Currently I am using something like:
num_bytes / (MBYTES_PER_SEC * time)
where I have arbitrarally picked MBYTES_PER_SEC to be 1 MegaBytes / sec.
Given a 10MBit (1.25 MByte) channel, subtracting overhead for preambles,
CRC bytes, collisions & the minimum time between packets, I came up
with the VERY approximate # of 1 MByte/sec.
I know that 1 Meg is a VERY magic number because there are many
different media that the data can be traveling across which will
effect this magic number.
So, back to the question. Is this a good approximation? What
values for MBYTES_PER_SEC should be used for the various medias,
i.e. think wire, thin wire, broadband, fiber, etc.?
I don't want to worry about distances between stations and the
media between them, just a rough approximation.
I have looked at the Blue Book to try to find a "nice" method
for computing network loads, but I am having a hard time trying
to come up with a better equation. Any help anyone could supply
me (as far a clarification of what the Blue Book is saying and
how to use that information) would be greatly appreciated.
Thanks,
--
@__________@ W. Tait Cyrus (505) 277-0806
/| /| University of New Mexico
/ | / | Dept of Electrical & Computer Engineering
@__|_______@ | Parallel Processing Research Group (PPRG)
| | | | UNM/LANL Hypercube Project
| | hc | | Albuquerque, New Mexico 87131
| @.......|..@
| / | / e-mail:
@/_________@/ cyrus@hc.dspo.govdonegan@stanton.TCC.COM (Steven P. Donegan) (04/06/88)
In article <23564@hi.unm.edu>, cyrus@hi.unm.edu (Tait Cyrus) writes: > > I am looking for an equation that roughly computes the theoretical > load on a network given only the following information: > 1) time over which the load is to be computed, > 2) the number of packets seen in that time, > 3) the number of bytes seen in that time. > > Currently I am using something like: > num_bytes / (MBYTES_PER_SEC * time) There is a fundamental question which must be asked before any answer to this question can be considered - what type of LAN are you trying to get bandwidth utilization numbers for? An Ethernet type LAN is extremely variable based on the number of 'stations' or 'nodes', differences in nodal load and collision detection timing, traffic rate per node etc. A token ring on the other hand can be very predictable in it's performance. Steve Donegan Sr. Telecommunications Analyst Western Digital Corp. donegan@stanton -- Steven P. Donegan Sr. Telecommunications Analyst Western Digital Corp.
henry@utzoo.uucp (Henry Spencer) (04/10/88)
> ... A token ring on the > other hand can be very predictable in it's performance. Until the token gets lost or corrupted, of course. -- "Noalias must go. This is | Henry Spencer @ U of Toronto Zoology non-negotiable." --DMR | {allegra,ihnp4,decvax,utai}!utzoo!henry