ExtremeSwitching (VSP)

 View Only
  • 1.  SLPP design ERS8k to VSP7k

    Posted 07-26-2018 10:23
    I'm currently looking at updating a customer deployment of SLPP within a datacentre. I have looked at the following design docs -

    Switch Clustering Best Practices NN48500-584 Version 8.1
    Switch Clustering using Split Multi-Link Trunking (SMLT) with VSP 9000, VSP 8000, VSP 7000, ERS 8600/8800, and ERS 5000 Technical Configuration Guide NN48500-518

    and also a seperate PDF from 2013 about loop prevention

    and I cannot find a clear solution for the deployment where we have -

    2 x ERS800 v7.2.26 IST cluster switches, linking to
    2 x VSP7000 10.4.4 IST cluster switches

    If this was a typical edge deployment I would simply enable SLPP on the core and SLPP-Guard on the edge but as the VSPs are connected via IST how does this affect the design? Should SLPP be enabled on the VSP VLANs along with the uplink ports? Would SLPP-guard still be used?

    there are some examples in the above docs but they don't really fit.

    I can provide a diagram of the setup.

  • 2.  RE: SLPP design ERS8k to VSP7k

    Posted 07-30-2018 09:32
    I would still use SLPP Guard on the edge. No matter the design, if you receive the SLPP packets on the edge ports, you kill the port ...

    For the 2 IST clusters, I would only enable SLPP on one of the 2 clusters. Ohterwise you will create to much overhead on the switches in my opinion.
    You can either choose to protect the core at all times and have the risk of losing the VSP7000 distribution cluster or configure the VSP 7000 cluster with SLPP to protect that cluster. Normally SLPP will kill the loop here and therefore also protect the CPU of the core, but there is no guarantee in that one.

    Hopefully this helps

  • 3.  RE: SLPP design ERS8k to VSP7k

    Posted 07-31-2018 09:12
    Hi Yves,

    thanks for your response. That pretty much matches our thoughts and the feedback we've had from Extreme. The main concern was around a daisy changed pair of VSP7ks, so we had VSP9k Core IST cluster> VSP7k ToR IST cluster > VSP7k IST cluster > server infrastructure.

    This was a full mesh deployment as well so the loop detection was key.

    The simple answer being, deploy at the edge.