BERT Testing with 3750 24 port switch

[wrgoulden]

I've got a situation where I have the following set up:

v100Mux > 3750 > Router >< Router < 3750 < v100Mux

Basically the mux encapsulates v.11 into Ethernet, passes it to the switch, the switch then passes it to the Router. The Router passes it to the other Router via an X.21 circuit, then the same on the otherside, simple.

My problem is when I place a BERT tester on the mux's and do an end to end test I receive bit errors and pattern slips. The interesting thing is if I replace the Cisco 3750's with a basic D-Link switch the BERT test passes. Any ideas?

The only information worth knowing about the D-Links is they are 10/100Mbps, use store and forward and 802.3x flow control. I've have configured everything on the 3750 I can think of and even left it blank, but it still fails the BERT. Any help is appreciated.

Thanks

Wayne

[johansens]

Hi Wayne,

Check to see if you have enabled QoS globally. If so, this may have negative impact on the perceived performance if you haven't done any more configuration.

If it's there, do a "no mls qos" and see if it helps.

Did it help?

[wrgoulden]

Thanks for your response, I wont be able to apply this command until the weekend as testing is underway but I will let you know.

I can say that a 3550 24 port switch passes fine, so I need to find out what the distinct differnce is in the switching.

Thanks

Wayne

[Georg Pauwen]

Hello Wayne,

which BERT (model) are you using, and does this happen with specific bit patterns ?

The flow control might somehow be involved here, the 3750 switches cannot send, only receive, the pause frames used in flow control. By default, flow control is set to off on the 3750 ports, you might want to try and set that to ´desired´ (´flowcontrol receive desired´ interface command), and see if that makes a difference.

Regards,

GP

[wrgoulden]

Hi GP,

Not near the BERT at minute but it's an Acterna, testing 64k X.21 DTE using a single bit pattern. The other end has a DB15 loopback plug attached. The test has been applied to all variants of switch, DLINK, NETGEAR and they pass.

Strangely enough I can now say the Cisco 3550 passes the test. I enabled the flow control feature on the 3750 but that made no difference.

As I mentioned in my post above, it would be interesting to know the differences in switching between the 3550 and 3750. Flowcontrol I beleive is on as default on a 3550 as with DLINK etc... but not the 3750.......

Thanks for you response

Wayne

[drussell]

The BERT error indications of bit errors and pattern slips lead me to think that you are losing packets in the path. Have you checked each of the ports involved for input/output errors or dropped packets? It could be as simple as a duplex mismatch.

[wrgoulden]

Hi David,

Interface statistics are clean, and I've hardcoded and autonegotiated the speed/duplex on both the end node and the switch. Used brand new cables, even used different 3750's.

I've got a test running over this weekend using 3550's and they have been running error free for 50+ hours now. I'm happy to stick with the 3550's but am puzzled over the 3750's and why I get bit errors within 20 mins.

I will continue to test with 3750's for my own piece of mind and exhaust the command structure.

Thanks for your response,

Regards,

Wayne