For fail over which is faster EAPS or a LAG/MLAG

[EtherMAN]

We have this question come up from time to time where we come off one switch and build redundant fiber paths to another stand alone switch. Most of the times we will build a simple EAPS ring between the connected ports and make the far switch the master and move on. In the last year we have started just creating a simple LACP lag between these switches using L3-L4 for our algorithm instead of an EAPS ring. Without me breaking out the test sets and setting up a lab which one will fail over faster if a single link goes down? The LAG scenario or EAPS? For anyone that answers did you lab this up and test it?

Curious minds need to know 

[EtherMAN]

Thanks Olaf, I have a couple of JDSU's 5800's and may just have to set a test bed up then... We have some of each running right now and we have up till this year always built and Eaps ring instead of using lag but the guys seem to think LAG is faster and for sure a few steps simpler to configure. No time line of when I get this on our schedule but when we do I be more than happy to post our findings back ...

[AnonymousM]

When MLAG first came out, I did some tests on failover times between MLAG and EAPS using VoIP and Video Streaming. I did not have access to proper testing equipment, so I can't quantify it in milliseconds, but the impression was that EAPS is a little bit faster than MLAG. In the phone calls there was nothing to hear and the session continued for both protocols, but in the video we could see a few pixels when using MLAG. Of course this is not representative and was in a very simple and small environment. When using big rings and lot's of VLANs, EAPS is getting slower as well. As a rule of thumb I would say, EAPS was netween 50 and 100 ms, MLAG is maybe between 100 and 150 milliseconds. That was with 12.5 with X46G1. So in todays software with newer hardware revisions, it may be even faster.