Cisco WAAS-Global policy for VMware Vsphere and/or 3par replication

[Natalie Ramirez]

So, this is somewhat annoying that VMware Site Recovery Manager 5.0 does not seem to get much replication acceleration, mostly it is just Pass Through.  I have read a couple of Cisco marketing powerpoints that say WAAS will accelerate VMware.  But there are no Policies to that effect or configuration assistance.  So, vmware has a hundred or so connections in passthrough, all using port 44046, this web site here:

http://kb.vmware.com/selfservice/microsites/search.do?cmd=displayKC&externalId=1009562

Shows that this is the port used for Ongoing replication, and port 31031 is used for the initial replication.  So, I have two 674-8gb with inline cards.  One in main office, one at DR office, both running 5.0.1.  The Lan ports on each side point to the LAN and the WAN interfaces on each end point to each other.  I have a layer 2 - 90Mbps link between the two locations, so "show cdp neighbor" shows the WAN interface connected.

My question is, is there a policy I should create for this data to be accelerated and stop being "PT Asymmetric"?

Second question, very similar to the first, 3par replication.  Same as above, but 3par uses port 5785 and the traffic shows up as "PT In Progress", there are only 6 or so connections in this state, but all are on this port 5785.

If I pull up the pretty graphs on the CM for this device, it shows a well distributed graph, but if I click the check box for "include Pass-Through" it turns all blue and says 100% of traffic is "other-traffic"  Since 99% of what goes between these two WAAS devices is VMware site recovery manager 5.0 and 3par replication, I would really like to find a fix for this.

[Natalie Ramirez]

ANSWER *******  SOLUTION  *******  ANSWER

I created 2 - Optimization Policy Rules for "WAAS-GLOBAL"

par3-rcopy, destination ports - 5785, 3491-3492, Application - Replication, TFO with DRE Adaptive and LZ

VMware-Replication, destination ports - 44046, 31031, Application - Replication, TFO with DRE Adaptive and LZ

Then I rebooted both WAAS devices and shut off the link for 20 minutes.  When I brought the link back online.  100% of data was accelerated, and 99% of the data was classified as "Replication" data.  I now get between 60% and 90% acceleration on this "Replication" traffic.  The final 1% is other data, remote desktop, ssl, citrix, sql, web...

Lessons learned:  The 3par and VMware keep TCP connections open forever, and once the traffic in that session is classified as something, "other traffic" or "Pass Through" it does not change until you reset the connection.  So, if you make any changes, you have to shutdown the link, and clear all TCP connections from the WAAS devices, then it will go to a different optimization rule. 

Final thoughts:  I am not completely sure that the Optimization policy rule "TFO with DRE Adaptive and LZ" is the BEST possible rule to use for this traffic.  If anyone has a better configuration for this traffic, I would really appreciate your input.