Theory ?

If you had 1000 identical windows host computers , all running identical
topology ( say the same ethernet ) , and a class A isp address assigned to
the network : would there still be a reason to subnet into network segments
( given the server is windows also ) ?

Would performance in terms of excessive broadcasts slow down the network ?

No wonder MCSEs make $ , it is almost overwhealming what must be known .

In contrast , Electronics was simple .

Thanks Much , Tim

Herb
Sun Feb 01 18:39:09 CST 2004

"Tim Kettring" <tim6kettring@e-garfield.com> wrote in message
news:bvju21$t98v5$1@ID-212626.news.uni-berlin.de...
> If you had 1000 identical windows host computers , all running identical
> topology ( say the same ethernet ) , and a class A isp address assigned to
> the network : would there still be a reason to subnet into network
segments
> ( given the server is windows also ) ?

Yes. The above network would likely perform ok but not as well as other
designs
-- depending on the amount of traffic each station was trasmitting.

> Would performance in terms of excessive broadcasts slow down the network ?

Probably somewhat of an effect but worse would be the sharing of the actual
bandwidth to data.

Were you to split the network with an (old style) bridge, you would still
have the
broadcast issue -- which would likely be acceptable with only 1000
stations -- but
then you would have multiple nets for data to be transmitted concurrently.

However, if all the "servers" were on one segment, you would be back in the
situation
where all the "clients" bottleneck on the server segment.

Consider the advantage of organizing the physical connections to match your
"business"
with "accounting clients next to accounting server", "engineering clients
next to engineering
servers", etc.

Thus the motivation for the new style VLAN switch where the connectivity
device maps
the physicial network to the business network.

> No wonder MCSEs make $ , it is almost overwhealming what must be known .

The real trick is learning how to THINK, how to TROUBLESHOOT, and how to
learn
new material rapidly.

Similar to "Moore's law", about every 2 years one half of your knowledge
will be obsolete,
but your ability to apply what you do know and to seek and learn what you
don't know, can
never be taken from you.

BTW, if I recall correctly the "record" for PRODUCTION broadcast domains was
something
like 22,000 (many segments with bridges) at IBM, DEC (remember them) was
second at
about 16,000, and the company I worked for was ONE of the larger one's at
about 9,000
nodes on a single broadcast domain.

Then Cisco invented "cheap" routers -- and the rest was history.

(Cheap meant about $25,000-$40,000 but his was in the days when a
mini-computer had to
do this job and these likely STARTED at $100,000 and could easily go to a
million.)


--
Herb Martin

Tim
Mon Feb 02 11:29:41 CST 2004

Thanks Much for the info :-)


"Herb Martin" <news@LearnQuick.com> wrote in message
news:%23CQW6TS6DHA.2696@TK2MSFTNGP09.phx.gbl...
> "Tim Kettring" <tim6kettring@e-garfield.com> wrote in message
> news:bvju21$t98v5$1@ID-212626.news.uni-berlin.de...
> > If you had 1000 identical windows host computers , all running identical
> > topology ( say the same ethernet ) , and a class A isp address assigned
to
> > the network : would there still be a reason to subnet into network
> segments
> > ( given the server is windows also ) ?
>
> Yes. The above network would likely perform ok but not as well as other
> designs
> -- depending on the amount of traffic each station was trasmitting.
>
> > Would performance in terms of excessive broadcasts slow down the network
?
>
> Probably somewhat of an effect but worse would be the sharing of the
actual
> bandwidth to data.
>
> Were you to split the network with an (old style) bridge, you would still
> have the
> broadcast issue -- which would likely be acceptable with only 1000
> stations -- but
> then you would have multiple nets for data to be transmitted concurrently.
>
> However, if all the "servers" were on one segment, you would be back in
the
> situation
> where all the "clients" bottleneck on the server segment.
>
> Consider the advantage of organizing the physical connections to match
your
> "business"
> with "accounting clients next to accounting server", "engineering clients
> next to engineering
> servers", etc.
>
> Thus the motivation for the new style VLAN switch where the connectivity
> device maps
> the physicial network to the business network.
>
> > No wonder MCSEs make $ , it is almost overwhealming what must be known .
>
> The real trick is learning how to THINK, how to TROUBLESHOOT, and how to
> learn
> new material rapidly.
>
> Similar to "Moore's law", about every 2 years one half of your knowledge
> will be obsolete,
> but your ability to apply what you do know and to seek and learn what you
> don't know, can
> never be taken from you.
>
> BTW, if I recall correctly the "record" for PRODUCTION broadcast domains
was
> something
> like 22,000 (many segments with bridges) at IBM, DEC (remember them) was
> second at
> about 16,000, and the company I worked for was ONE of the larger one's at
> about 9,000
> nodes on a single broadcast domain.
>
> Then Cisco invented "cheap" routers -- and the rest was history.
>
> (Cheap meant about $25,000-$40,000 but his was in the days when a
> mini-computer had to
> do this job and these likely STARTED at $100,000 and could easily go to a
> million.)
>
>
> --
> Herb Martin
>
>