WCCP v2 Multiple interfaces and TCP breaking traffic

vipul.vajshi · ‎08-13-2007

Hi,

I am having issues with trying to configure WCCP on a 3750 and 4948, getting the same issues at both ends.

I have 3 interfaces where I need to intercept traffic.

i.e

Branch Office - 3750 12.2(37)SE

int g1/0/1

ip wccp 61 redirect in

WAN Link1

int g2/0/1

ip wccp 61 redirect in

WAN Link2 - This goes off to another site where we do the same as the below device, we have a triangle.

int vlan 300

USER DATA NETWORK

ip wccp 62 redirect in

WAN EDGE DEVIC - 4948 - 12.2(31)SG

int g1/1

ip wccp 61 redirect in

WAN Link

int g1/47

ip wccp 62 redirect in

CORE Link1

int g1/48

ip wccp 62 redirect in

CORE Link2

TCP traffic breaks when we configure WCCP like the above, is that a valid config, if so why is it breaking tcp traffic, though we don't have issues with icmp as expected.

Thanks,

Vip

Zach Seils · ‎08-17-2007

Vip,

The interface configurations look correct. Is all TCP traffic affected? It would be beneficial to see the following output from the switches:

show ip wccp

show ip wccp 61 detail

show ip wccp 62 detail

In addition, please provide the WAE configuration from each site.

Thanks,

Zach

vipul.vajshi · ‎08-20-2007

Hi Zach,

Thanks for getting back to me.

As we had this issue, I have had to take the WCCP config off the switches, and yes all tcp traffic was being black holed.

As long as I did not have 'ip wccp 61/2 redirect in' running on more than one interface, it worked fine.

So I don't know if it was wccp that was causing the issue.

Also the WAEs are running on different versions.

The configs are attachted.

Thanks,

Vip

Zach Seils · ‎08-20-2007

Vip,

A few suggestions:

- Disable the EPM Adaptor

- Disable WCCP Slow Start

- Use the default WCCP mask

Can you also check syslog.txt for any errors. Specifically you should look for any indication of a redirection loop.

Zach

tblancha · ‎08-21-2007

So, g1/0/1 and gig2/0/1 are the WAN links at core? If so, it might be that ip wccp 61 rediret in goes on the LAN interfaces where the clients/servers reside. Then ip wccp 62 redirect in goes on the interfaces that represent the WAN interfaces. Is it possible these are switched in your config?

samerg · ‎08-28-2007

Hi,

have you been able to solve your problem? Looking at the configs, I'd suggest the following:

1) On the switches, use 61 redirect in on the LAN side and 62 redirect in on the WAN side.

2) Upgrade your WAE's to 4.0.11 or 4.0.13.

3) On the WAE's, disable wccp slowstart.

4) On the WAE's, use this wccp config:

wccp tcp-promiscous router-list-num 1 l2-redirect mask-assign

Regards,

Sascha

vipul.vajshi · ‎08-30-2007

Hi Sascha/Everyone else :D,

I have been away and we had a tight deadline to get this in, I have upgraded the unit to 4.0.11.34 and also now using PBR and this has done the trick.

Though I am having one problem, in the RDP session we are not seeing any reduction or performance.

I have moved the policy to the top, I also see the connection:

172.26.9.231:4799 172.23.3.22:3389 28455 00:14:5e:95:62:6f F,F,F,F

And the moment it is configured as such:

MS-Terminal-Services Remote-Desktop Optimize(DRE,LZ),Accelerate(CifsAccelerator)

We are seeing performance improvements in other areas, CIFs, FTP, etc.

Thanks,

Vip

Zach Seils · ‎08-31-2007

Vip,

The default policy for RDP traffic is TFO only, since the packets are usually small and don't benefit much from compression. Depending on the WAN characteristics you are testing with, you may or may not see a marked improvement.

I also noticed that your policy includes the CifsAccelerator, which is not valid for RDP traffic. The CifsAccelerator should only be used for CIFS traffic.

Thanks,

Zach

vipul.vajshi · ‎09-03-2007

Hi Zach,

Thanks, I was just messing around with CifsAccelerator and seeing if I was going to get any improvement, I know it is for WAFS, but trying to see if I can get any improvments.

:D

We are running this on a 2MB MPLS circuit (MPLS circuit provided by WAN provider) that goes out to Australia.

So there it is unlikely that we will see any improvement?

Though one thing for sure that is odd is the statistics report shows this, I think there might be a bug in the reporting tool, as when I check for each application, there is reduction and optimization.

Thanks,

Vip

Total bytes saved: 0 KB

Percent reduction (incl. pass-through) provided by this device: 0 %

Percent reduction (excl. pass-through) provided by this device: 0 %

Effective capacity provided by this device: 1 X

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Application Name Application Traffic (%) Effective Capacity Reduction

Total (%) Optimized Only (%)

Backup 0 1X 0 0

Content-Management 0 1X 0 0

Directory-Services 0 1.6X 37 37

Email-and-Messaging 3 1.3X 25 25

Zach Seils · ‎09-04-2007

Vip,

What is the latency of the link?

Zach

vipul.vajshi · ‎09-05-2007

Hi Zach,

Latency is about 330ms.

Regards,

Vip

Zach Seils · ‎09-07-2007

Your BDP is ~81KB, and you have the buffers set to 512, so you should be ok. I suspect in your case the RDP performance is not bandwidth constrained, but rather latency bound.

Zach

vipul.vajshi · ‎10-18-2007

Hi Zach,

Thanks for getting back to me on this, case closed.

:D

Regards,

Vip