Load balancing in Nortel cluster.

[Peter Nguyen]

Hi All,

How do traffic load balancing in Nortel 8600-5500 switch cluster?  

My triangle cluster comprise of:

• Two 8600s with 8 of 1 Gb MLT-IST together.
• Stack of 5510 with 8 Gb ports SMLT 4 each to the two 8600s
  

 “Show mlt ist info” and “show mlt smlt info” indicate IST and SMLT are working.
 “Show ports info vlacp” indicates all ports in the cluster are up and functioning.

The 5500 stacks connects to a VMFarm.  MRTG showed only one MLT-IST and one SMLT port saturates with at 95% bandwidth and the other ports in the cluster are doing almost nothing.  

Software for my 8600 is 4.1.0.0, and my 5500 is 5.0.0.4.  My 8600’s CPU is 8691SF; Gb blade is 8616GTE; no R-modules.  

Any ideas, suggestions are appreciated.

Regards,
Peter.

[Michael McNamara]

In a MLT the switch hashes the MAC address of the source and destination (this is configurable on some switches and software releases) to determine which port to use in forwarding the frame/packet. This guarantees that traffic to/from the same source will flow over the same link and avoid any sequencing issues. The algorithm doesn't generally guarantee a 50/50 dispersion of traffic but it's usually not far off. With that said if you have a single TCP conversation consuming large amounts of bandwidth then MRTG will reflect this when graphing the utilization of the ports in the MLT.

If you have a single VMWare host generating all that traffic then this is normal behavior.

There's a much better explanation in one of the technical configuration guides. In the later software releases you can change the algorithm if you wish... I would advise caution and make sure you test extensively before making any changes.

I'll see if I can find that reference for you.

Cheers!

[Peter Nguyen]

Michael,

Thank you for your clear and detail explanation.  This is my first cluster builds to support VM farm; many unknown issues to both server administrators and network infrastructure groups.  Our SE is more confusing than helping.  My configuration based on TCG NN48500-542 “Resilient Data Center Solutions for VMware ESX Servers Technical. “  If you have better document, please let me know.

Best Regards,

Peter.