0

following secario: A MX480 with MPC3e 3D NG Linecards.

I have a new Layer2 Access provider that is providing Layer2 services between 100 and 1000MBit/s. My connection zu this services are a 10GE E-NNI interface.

This provider is policing the traffic hard to the service bandwidth, allowing 60kB of burst, independent of service bandwidth.

Here comes my struggle: The shaper on the MX Platform has a minimum burst of about 1,8ms, so you can't implement a real "zero Burst" shaper here. There are ways of reducing this minimum bust by reducing the queues on the FPC ... i'm in contact with JTAC about this. But i don't think that this will solve my issue.

With this 1,8ms i'm able to stay within my provider's contract for services up to about 500MBit/s, but on service bandwidths of 600MBit/s and above the Juniper is not able to stay within the 60kB due to the minimum Bust of 1,8ms.

In my understanding i have to switch from a CIR based approach to a PIR based where i configure a CIR that is quite below my provider's CIR, and have to tune my burst limits up until a point where i still comply to the limit of my provider's PIR.

On i Cisco Box i would switch from a "shape average" configuration to a "shape peak" configuration with something like an average rate of 50% and a target rate of 100% of my provider's CIR.

Can anybody tell me the approach to solve this on a Juniper? I'm missing some information about the token refill rate of a Juniper shaper, how often can i access these Burst tokens i configured?

Thanks for your support and any comments here!

regards //Andreas

  • 2
    You'll never be able to achieve a high utilization rate before experiencing undesired packet loss with 60KB burst on 100Mb/s+ services. It's totally unrealistic. Worse, your efforts to try to do so will almost always move packet loss from the provider network to your own network, or even worse, result in more delay buffering, without really solving the problem. Consuming services like these involve compromises. Note you should probably be using the enhanced queueing cards for this type of concentrator configuration, too. $0.02. – Jeff Wheeler May 1 at 22:27
  • thanks for your comment. I'm in contact withe the product engineering of my provider to clarify if it is correct that we see a maximum burst of 60kB along all service bandwidths and if we are able to change this. But, dealing with large carriers always takes time. – Andreas Schaefer May 2 at 6:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.