What's the plan for the network reorganization?

Note: in these figures, network numbers without a P are 152.2.### and with a P are 192.168.### (private).

The OLD networks used to look like this:

The basic plan now is to have two separate networks. One network, the experimental network, has LOTS of machines (~12) so we can do long-lived experiments with heavy traffic loads (hopefully without overloading the traffic generators in terms of processes, file handles, etc.). The other network, the test network, has a significantly smaller number of traffic machines (6). This network allows us to test traffic generation tools, router code, etc. and still generate most of the extreme cases (e.g. overloading the router) without tying up the experimental network.

The new network figures are below. They are now in a single powerpoint (for updating) or PDF document.:

What changes between the old network and the new one?

Network hardware

All of the hubs and switches are now labelled in the lab and on these diagrams in an effort to better manage the configuration.

The test network will use the catalyst 5000 that was our department connection and the Cabletron SmartSwitch 2200 for the end-system networks. The experimental network will use separate VLANs on the "experimental" catalyst 5000. The catalysyt will be reorganized into 2 VLANS (instead of 4). I chose this allocation of the different switches because it offers better symmetry in the experimental network and continues to use the same technology for experiments. Since we don't know if the 2200 has any odd behavior I wanted it in the "test" network.


First, let's cover the experimental network. Some machines simply move from the 290 network (or from standby status) to the experimental network:

and some 290/standby machines will change names to replace DiRT machines: Also, one machine from the closet becomes a traffic generator and takes the name of one of the machines returning to the MSL:

On the test network, a couple of machines move into service as traffic generators:

Other machines are returned to their original owners. Also, a new machine acts as the gateway to the 290 network and the former gateway goes into standby (to give the 290 students time to package their tools, etc. for us in January.):

Finally, 4 machines come into service from the closet to serve as machines for Prasun's 242 class. (perhaps bennett will also be used for this if Prasun needs 5 machines?) All of these machines are new even though they will be using the names of some of the 290 machines that have changed names. NOTE: I chose these machines since the time that they will be used by the 242 class matches the loaner period. Using other machines for this would mean more reconfig at the end of the spring.

The production machines will continue to serve as production machines plus additional functions as noted:

Some notes and rationale:

I used Bert's machines for 242 since the time they are on loan matches the time Prasun will need them. (Using our machines for 242 or Bert's machines in our networks would mean reconfiguration at the end of the semester for us.)

I tried to keep names intact in the "DiRT" network and only add new names in the hopes of reducing the amount of script and configuration file changes I and others would have to make.

I believe that using private network numbers (192.168.*.*) for the "bottleneck" networks might allow us to get by without acquiring any new subnets.

I left krishnan with multiple network connections in the experimental network on the assumption Michele will continue her work. Perhaps it should be in the test network for a while?

I didn't count wilee as a traffic machine (although it's in the 138 network) and figured it may move to the test network if it continues to be used for Prop.Share.

Topsecret and dot will eventually swap names so that the faster machine (dot) can be used as a 2.2.8 production machine when Mark leaves.

Other DiRT documents
Author: Mark Parris
Last updated: December 15, 1998