Showing results for 
Search instead for 
Did you mean: 

WCCP performance overhead


Hi All


We have implemented WCCP-GRE from some of our remote wan routers to proxies in  the Data Center. We’ve followed best practice guidelines, bug scrubs, TAC review etc – so nothing stands out from a configuration standpoint. However, I have not had much luck in getting insight into the following and so posting here


I’d think there would be ‘some inherent’ processing/overhead/delay from an end user perspective when the traffic is redirected using WCCP-GRE vs explicit proxy. Basic laws on encap/decap etc would cause that delay. However, is there any reference published on what that delay percentage is. For instance, is WCCP about 2 times (or n times) slower than explicit proxy configuration on browser ?


Our users have been gauging performance based on and reporting 10x times performance degradation numbers using WCCP. But, speedtest is not a true representation of actual user experience. However, am not able to quantify what is a typical % when using WCCP vs explicit.


Thanks much

2 Replies 2



First of all : I don't think that you would ever find such a "slowness" factor, as it depends upon many things.

Basically there are several thing that can slow down your traffic :

1) WCCP can (inaccidentally) be handled in software on some platforms - this will slow down the router

2) You're adding both a GRE header and a WCCP header - beware of fragmentation


But to be honest - I've never seen such a setup, that you're describing :

You're redirecting WEB traffic across the WAN (truly I don't know the WAN delay), but normally you would place proxies close to the user and not in the datacenter, where you internet connection probably is terminated.

Consider what would happen if the proxy cannot handle the request (not necessarily a cache miss) - the packet would be returned to the originating router (across the WAN again for normal forwarding by the originating router), so you'll end up with traversing the WAN 4 times instead of two

And since many WEB pages consists of many elements (objects) delaying some of them might influence the entire response time.

Don't know whether this helps you, but I would place the WCCP in the Datacenter, close to the proxy.


Best regards

Finn Poulseb




Thank ya much. Appreciate the response.

We use bluecoat proxies and installing/maintaining those at 100+ remote sites is not an option. Egress to internet from all sites has to be through one of the regional DCs (in US, EMEA, APAC etc) by company policy.

Currently these users go to the regional DCs where the proxies are setup if using explicit mode. So, I had to mimic that using wccp-gre where possible. At one regional site though,  yes – I did implement L2-WCCP with proxy in the DC where the local users are.


But, in general, even with proxy local to the user - plain using WCCP seems to have a hit. Some amount expected due to the encap/decap - just trying to gauge that to set expectations. 


 #1 – yes – monitoring key health via nnmi for slow downs.

#2 that you suggested, agreed and we’ve been watching that.  We’ve been playing with the MTU settings with marked results.






Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Recognize Your Peers