WAAS - Failure Scenarios

[nathan.deane]

Hi,

In the situation where you have two locations connected by a leased line, one site is the DC and one the remote site. Both sites use WCCP redirection to the WAE which is directly off the WAN router. If the remote site WAE fails what is

a) the consequence for the optimised traffic? does the DC WAE notice that the TCP session has been terminated and ceases to optimise traffic to this location?

b) what is the estimated time before the DC WAE notices that the remote site has gone down? I am assuming all TCP sessions will be terminated and will need to setup again but this time from end to end?

Thanks

[dstolt]

a) The client or server behind the failed WAE would get traffic with inflated sequence numbers and would reset the traffic.

b) Immediately upon new connections via autodiscovery and also on existing conections via the reset (in a above), new TCP sessions would go through setup and autodiscovery again. If there is no WCCP cluster, they would be non-optimized.

Hope that helps,

Dan

[nathan.deane]

thanks, that explains it.

[nathan.deane]

one last point, when you say that the failover would be immediate, there must be a 1 or 2 sec lag whilst tcp sessions are rest and commence again?

are there any documented expected recovery times by cisco?

I know its a pedantic question as the whole inner workings of TCP will make it almost seamless.

Thanks

[s.uslay]

I have tested this in our prod network before deployment, central site WAAS with WCCP to a WAN router, branch WAAS is inline. Reloading of remote WAAS takes 40 seconds.When I turn of the remote WAAS it takes exactly same 40 seconds for teh device to fall into the wire.Some applications recover, some fail depends application.

Serhat

[Zach Seils]

Sounds like there is something else at play here. Perhaps spanning-tree? Do you have the switchport connected to the router (through the inline module) configured for portfast?

Zach