Question about redundancy on multiple WAEs

snakayama · ‎05-23-2007

Hi everyone,

I have the question about redundancy configuration of multiple WAEs.

I understand redundancy and load balancing of multiple WAEs can be achieved by WCCPv2 function (service group), and necessary commands on the router(s) are;

ip wccp 61 group-address x.x.x.x

ip wccp 62 group-address x.x.x.x

x.x.x.x is multicast address that is used to communicate between router(s) and WAEs in the WCCP service group and is used by router(s) and elected lead WAE to determine which WAE should receive redirected traffic.

I think redundancy and load balancing on multiple WAEs closely related because WCCPv2 in case of redundancy works like active/active failover and does not work like active/standby failover which means all WAEs under same service group always handle/receive the redirected traffic at a certain rate determined by load balancing mechanism, and each WAE does not have another WAE's cache information.

In other words, multiple WAEs have to be configured under the same service group to achieve redundancy which means it is necessary that multiple WAEs are controlled under load balancing and have to need to configure load balancing to get redundancy worked correctly.

I think redundancy mechanism of WAE uses WCCPv2 keepalive to determine the member WAE runs normally of not.

If one WAE failed any reason, another WAE is elected and begin to handle subsequent traffic in stead by using load balancing mechanism.

Is my understanding correct ?

And Do I need to configure just commands below on router(s) ?

ip wccp 61 group-address x.x.x.x

ip wccp 62 group-address x.x.x.x

or I need to configure other command to configure redundant WAE environment.

Your information would be appreciated.

Best regards,

Zach Seils · ‎05-27-2007

Nakayama-san,

Your understanding is correct. Note that the use of the 'group-address' argument is not required. The default unicast method should suffice, and is what is more commonly deployed.

In addition to the global service commands (which you have listed), you will also need a per-interface command for each interface where you want to perform WCCP interception:

!

interface FastEthernet0/0.101

encapsulation dot1Q 101

ip address 10.88.80.1 255.255.255.128

no ip redirects

ip wccp 61 redirect in

!

Thanks,

Zach

snakayama · ‎06-01-2007

Seils-san,

Thank you very much for your reply and sorry for my late response.

I have an additional question about "The default unicast method" you wrote.

I guess "The default unicast method" is as follows;

In redundant environment, each WAE must have a list of routers by configuring WCCP router list on each WAE. Through WCCP version 2 communication, each WAE announces its presence to each router on the rotuer list. So all of routers in the service group can have consistent view of WAEs.

And all of routers reply to all WAEs with their view of WAEs in the service group. So all of WAEs in the service group can have consistent view of routers.

After the view is consistent across all of the routers and WAEs, all of routers can identify the lead WAE which is one seen by all routers.

So all routers can just redirect traffic to lead WAE and lead WAE determines how traffic should be allocated across WAEs in the service group which means that all routers just know the unicast IP address of lead WAE.

Is my idea correct ?

Your comments would be appreciated.

Best regards,

kgovind79 · ‎06-01-2007

Snakayama,

Your idea is partially correct. From section in WCCP2 draft:

4.9 Electing the Designated Web-cache

The designated web-cache must be receiving a WCCP2_I_SEE_YOU message from every router in the Service Group.

Election of the designated web-cache is not part of the WCCPprotocol. However it is recommended that the web-cache with the lowest

IP address is selected as designated web-cache for a Service Group.

So all the routers would redirect the packets to all the WAEs according to the hash assignment specified in the lead cache in the REDIRECT_ASSIGN message - which is sent ONLY by the lead/designated WAE. The lead WAE is just like another WAE but just controls the distribution of traffic across multiple WAEs, and the router complies with this distribution and assigns the buckets accordingly for all the WAEs in the service group.

Zach Seils · ‎06-01-2007

Nakayama-san,

You are mostly correct, except that the lead WAE actually tells the routers how to distribute to load. So redirected traffic goes directly to the assigned WAE.

Thanks,

Zach

snakayama · ‎06-04-2007

kgovind79 and Seils-san

Thank you very much for your reply and explanation.

I understand how routers identify lead/designated WAE, how routers would redirect the packets to all WAEs and the meaning of "Unicast method".

Thank you very much for your assistance.

Best regards,