Node Group Latency/Bandwidth Guidance

[paul]

Typically I only do node groups when the PSNs are on the same subnet, but now that MAR cache distribution is supported between node group members I am wondering what the basic guidelines for latency and bandwidth consumption for node group synchronization.  For now, I am working on a small customer with one node in their primary site and one node in their DR site.  I am upgrading them to 2.3 and they are using MAR + User Auth for their wired environment.

Or is just a bad idea to attempt this across a WAN.

[Craig Hyps · Accepted solution]

Guidance is still the same. (See ISE Latency and Bandwidth Calculators). We assume LAN links (think GE or better) between nodes in same node group.  We do not test with less.  In general, all PSNs in same Campus LAN should be part of same node group.  Unless considering dark fiber connects between DCs, the separate DCs or any other WAN connection should be grounds for separate node groups.

[paul]

Thx Craig. Yeah I didn’t see anything on the calculator specific to node group bandwidth and latency as one of its assumption is node groups are contained to one location. I knew the general guidance for a deployment as a whole.

I figured doing inter-datacenter node groups would still not be recommended. I would have never entertained the idea of trying this, but MAR distribution is a nice sounding feature, but in practicality not going to help for smaller deployments split between two datacenters. I was hoping. ☺

Thanks for the quick feedback.

Paul Haferman

Office- 920.996.3011

Cell- 920.284.9250