Predicting Queue Throughput

It took some time to get here. CoS is a big subject, and the Trio capabilities are so broad in this regard that it can be overwhelming at first, even if you are already familiar with the general CoS processing and capabilities of Junos platforms. This section is designed to serve as a practical review of the key points and behaviors covered. The approach taken here is somewhat pop quiz-like to keep things fun.

If you find yourself surprised at some of the answers, read back over the last 100 pages or so. The truth is out there.

CoS is one of those tricky subjects; the kind where everything is working fine, and so, feeling bored, you decide to make a small tuning adjustment, for example explicitly assigning an excess rate to a queue that is equal to the one it has already been using through default inheritance. Yes, a very minor, very small, seemingly innocuous change. Yet boom! Suddenly all sorts of behaviors change and you are once again eternally thankful for the rollback feature of Junos. It’s hard to test CoS in a live network for numerous reasons that are so obvious they need not be enumerated here. The takeaway is that you should test and model the behavior of any proposed CoS change in a lab setting to make sure you understand all behavior changes before rolling the proposed change into production. Small changes really can have big impacts, and these are often unanticipated, as the material in this section is intended to demonstrate.

Where to Start? ...

Get Juniper MX Series now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.