WAAS Peering in a redundant setup

Unanswered Question
Jun 10th, 2009
User Badges:


Any idea how this autodiscovery works in a redundant setup? where i have 2 WAE in the Head office and 2 WAE in the DR site. How will the 2 WAE in the HO peer with the 2 WAE int he remote site?


Autodiscovery of WAAS Devices

Cisco WAAS includes an autodiscovery feature that enables WAEs to automatically locate peer WAEs on your network. After autodiscovering a peer device, the WAEs can terminate and separate the LAN-to-WAN TCP connections and add a buffering layer to resolve the differing speeds. Once a WAE establishes a connection to a peer WAE, the two devices can establish an optimized link for TCP traffic, or pass the traffic through as unoptimized.

The autodiscovery of peer WAAS devices is achieved using proprietary TCP options. These TCP options are only recognized and understood by WAAS devices and are ignored by non-WAAS devices.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
jujouber Wed, 06/10/2009 - 15:29
User Badges:
  • Cisco Employee,

When a WAE in the Data Center receives a SYN packet that is marked with the TCP Option 0x21, it knows there is another WAE on the Branch side and marks the returning SYN_ACK with the same TCP option. This is auto-discovery in a nutshell.

The WAE's are then able to peer and optimize traffic. If the SYN_ACK does not pass through the WAE in the HO, then it returns to the Branch unmarked and auto-discovery fails due to an asymmetric flow and no optimization is possible.

As for choosing which WAE to peer with in the Core, in the 4.1.x code it will depending on the traffic interception method used, ie WCCP or GRE. Which ever WAE receives the packet will mark the packet as described.

In the WAAS 4.0.x code, the 4050 tunnel that was built between the Edge and Core of CIFS optimization worked differently, it was a random determination.

mta2891974 Thu, 06/11/2009 - 05:31
User Badges:

Thank you for your reply.

I'm attaching the topology.

Actually the 2 sites will be connected by redundant routers and links and EIGRP load balancing between these 2 links. So asymetric routing is there: outgoing traffic might pass through WEA1 and WAE3 and return traffic WAE3 (or WAE4?) and WAE2!

In order to avoid asymetric situation we want to design WCCP as ACTIVE/STANDBY by adjusting the "weight" field for WCCP load balancing. I have the following questions:

1- if we choose a weight of 100 for WAE1 and WAE3 and weight of 10000 for WAE2 and WAE4. Does this mean that 100% of the traffic will flow between WAE1 and WAE3? (and no traffic between WAE2 and WAE4?) and in case of a WAE failure (say WAE1 is down), WAE2 will take 100% of the traffic load?

2- Traffic flow: we know that WCCP will load balance by default based on the source ip and then the weight: this is for incoming traffic passing through the LAN sub-interface of the router (where we have " ip wccp 61 redirect in") what about the return traffic from the remote site? return traffic entering the WAN interface of the router (where "ip wccp 62 redirect in" is configured) does the weight applies here too?

3- Egress traffic (return traffic from WAE) to the router will choose the original router that sent him the traffic since we will be running the latest WAE version and configuring GRE method. is that correct?

My purpose is to implement WAE redundancy with automatic failover without load balancing at all and make sure that always WAE1 and WAE3 are active and WAE2 and WAE4 are standy and will take all the traffic only in case of failure.



Are the redundant WAEs in the same subnet in the Data Center? If so you can rely on WCCP hash/mask to send the traffic to the correct WAE everytime. I have this running in a similar setup where two WAN routers share a common WAE subnet and redirect traffic coming in two seperate WAN links to the WAEs. It works perfectly. I also use GLBP on the WAE subnet instead of HSRP to provide additional load balancing of traffic.


This Discussion