I have a situation where I have a medium sized Frame Relay network with OSPF as the routing protocol. It is a heirarchical partial mesh, with two central data centers, a number of primary sites and a larger number of secondary sites. My question has to do with how folks choose to use BANDWIDTH commands on the interfaces.
To illustrate the problem is when you have a router with an access rate of, say, 384k, and four PVCs. Two of the PVCs go to the data centers, and two go to other locations, one with 56k access, the other with 384k access. All four PVCs are 16k CIR.
Now, my question is: How do *you* personnaly choose to assign banwidth statements? I am fairly certain that there is no "right answer" to the question, as I have been unable to find any comments on this on CCO or elsewhere. Maybe opinions are the best way to go. Here's some ideas I've thought of:
-Just assign the Bandwidth statement to the main Serial interface, but don't add the statment to subinterfaces. All the bandwidth statement would do is indicate the access speed to the FR network.
-Calculate a proportion for each PVC and divide it into the access rate that the router has to the FR network
-Assign the bandwidth statement based on the max pipe speed possible between the near and far end of the PVC
-Assign the bandwidth statement simply based on the PVC's CIR.
-Use the bandwidth statement to "weigh" a PVC, weighing the data centers heaviest, primary sites less, and secondary sites the least. The bandwidth statement wouldn't indicate bandwidth at all in this case.
What other ways do you use to indicate the link speed, reliability and other preference mechanisms that should go into the routing protocol's link metric calculating routines?
Even if your response is not scientific, or perhaps you "just to it this way" and never thought about it alot, I'd like to hear from you. I am just looking for thoughts on how real people do this on their real networks.
Thanks!
Jim
