Bandwidth and Latency requirement for WAAS usefulness

[fast5658659]

Hi All,

        What is the recommended bandwidth and latency requirement between two site for WAAS for work effectively?

[Michael Korenbaum · Accepted solution]

Hi,

The ideal WAN environments to use WAAS in would be where the latency is 20msec or greater.  Once you get below 20 msec rtt you may want to disable the L7 accelerators (e.g. CIFS, HTTP, MAPI, etc.).  Not to say that you can not use WAAS if the latency is less than 20 msec, but you will want to do thorough testing with an accurate user profile to ensure that the WAAS overhead does not negate the savings it provides (e.g. perceived performance may be the same as without WAAS or possibly worse, even though WAN bandwidth utilization was low).  As for bandwidth you can have up to a 1Gbps pipe; I've personally only seen WAAS used in OC-3 WANs or less but I'm sure there are other deployments out there with greater WAN pipes effectively using WAAS (assuming they have a high latency).

Remember, the closer you get to LAN like latency and speeds the less of a need there is for any WAN optimization technology.  For example, if you had a OC-3 WAN with 3 msec of delay that probably would not be a great candidate for WAN optimization.  Even if you were to deploy WAAS in that environment and only use L4 optimization you probably wouldn't see much of a benefit from an end user performance perspective.

Hope this helps,

Mike Korenbaum

Cisco WAAS PDI Help Desk

http://www.cisco.com/go/pdihelpdesk

P.S.  If this answers your question please mark it as such; thanks.

[Tod Larson]

Michael,

Can I configure WAAS to only optimize connections that have RTT > 20 ms?   I'v got a dozen sites all with <20 ms of latency between each other.  I've also got another dozen sites within 20 ms of each other.  Howerver, these two group of sites are 100 ms apart.  All the sites have WAAS.  Can I tell WAAS to only optimize traffic to and from the sites that are 100ms apart while simply passing through traffic destined for the local sites?

[Michael Korenbaum]

Tod,

There is no way to restrict optimization based on latency.  However, since you would know the IP ranges involved you could do a few things.

1)  If you just don't want to redirect the traffic period configure WCCP redirect ACLs to prevent the traffic from even being sent to the WAEs

2)  If you are using Inline Interception you can use the Interception ACL feature introduced in 4.2

          http://www.cisco.com/en/US/docs/app_ntwk_services/waas/waas/v421/configuration/guide/traffic.html#wp1206910

3) If you just want to allow L4 optimization for these sites modify the default policy to permit only the ip ranges that you want to have L7 acceleration for, thereby implicitly denying these IP ranges from that policy and having the connection be handled by the default "other" policy.

     Or probably a better idea for tracking/reporting purposes; explicitly define new custom policies for these IP ranges to have L4 optimization only.

     For example, define a policy with source IP range and dest IP range and dest port of 445 or 139 and set this policy to TFO, DRE, LZ but don't utilize      CIFS Accelerator.    

Cheers,

Mike