From: Howard C. Berkowitz (hcb@gettcomm.com)
Date: Mon Jan 06 2003 - 14:24:24 GMT-3
At 10:32 AM -0600 1/6/03, Michael Snyder wrote:
>Let's create some Universal Topology layouts to match the produced labs.
>
>For example labs 1-7 run on a standard UT1.1 compliant rack
>
>Labs 7-12 run on a UT1.2 compliant rack.
>
>Etc.
>.
There's a philosophical question right off in dealing with lab
topologies, and I'll freely admit that I think we took the wrong
approach at Gett for the individual (but not the
instructor-led/bootcamp) racks.
One of the big issues is how much to control cost (and thus end user
pricing) versus how closely to mimic what's believed to be in the
real lab. We went in the latter direction, which had a big cost
impact when we swapped out all the 25xx's (other than terminal
servers, etc.) for 26xx/36xx, mostly for MPLS capability and also for
voice cards.
You can keep the cost lower by accepting the use of 2500s for most
issues, and perhaps having a smaller number of more expensive labs
for specific technologies. Now, we have a mixture. Other cost issues
deal with perception. Unless you are working with FE/GE and VLANs,
it really doesn't make that much difference, at the routing level, if
your LAN connections are Ethernet or TR. TR is now much cheaper, but
people are scared of practicing on it -- I think that's an unreal
fear.
Another big issue is to what extent to use "backbone" devices, such
as routers that have "proctor-controlled" configurations and are used
simply as route generators and test devices. You can come up with
single-pod configurations with a larger number of cheaper routers,
but it won't match the lab pod. How important is that?
Another consideration comes when you are setting up online services
with multiple pods. At that point, shared backbone equipment can
have massive economies of scale, and even do things that simply are
impossible in the 6 router/2 switch (well, 2 3550) labs, like come up
with complex AS paths.
There are different answers and approaches here, but I think that as
people put out labs, it's worth their mentioning their assumptions in
these areas.
Other things should be open to discussion. Do you really need two
3550s, or can you adequately do the 3550-specific things on one 3550,
and do the fancy general switching things with a 5000 (example)
connected to the 3550? What about study configurations when you have
bridging on a router working with one 3550?
.
This archive was generated by hypermail 2.1.4 : Sat Feb 01 2003 - 07:33:43 GMT-3