Friday, May 27, 2011

Just in time for Halloween

If you're using IP SLA tests to measure latency across your network, you may want to keep in mind that there is a caveat to using IP SLA.  Two reasons:
  1. QoS markings. If you are doing tests of different QoS queues all tests that are generated on that router will go out the default queue, even if marked for higher  priority.  So you use a shadow router that is within its trust boundary so that it will honor the QoS markings.
  2. Load on the Router.  IPSLA can add load to the router, on both the source and target.  So you can use the shadow router to act as the source or destination to reduce load on the WAN routers.  It’s not going to be a big deal if there’s only a handful of tests but if you’re doing an NxN combination of tests to between each site I would recommend a shadow router at each.
A shadow router?  Wait, i thought my Magic: The Gathering days were over after high school?!
A shadow router would simply be any router behind the WAN router, but within the QoS trust boundary.  That way any QoS markings on the test would be set on the shadow router, sent out the shadow router's default QoS queue, but honored by the WAN router and the rest of the network (depending on the config of the rest of the network).

Usually a shadow router isn't actually used for routing, but that will depend on your resources.  If you use a spare router to do your IP SLA, then the resources taken up by the IP SLA test aren't in competition with production resources.

Shout out to Zack Belcher who reminded me of these reasons to have a shadow router.