Brocade 3800 ISL over DWDM

[Stuart Walker]

Folks,
We have a couple of Brocade 3800's that we're trying to connect together(create ISL link) via a pair of Cisco 1500's(rebadged Inrange Spectrum). We're only talking about a distance of ~20km. Now, in order to make money, Brocade will "sell" you an Extended Fibre License. From what I can see, all it does is change certain parameters which cannot otherwise be changed - including bb_credits.

My supposition is this:

Without this Extended Fibre License, our ISL links SHOULD actually operate without a problem, but we may not be able to fully utilise the links. Mind you, I would have thought that the default bb_credits setting of 16 would be fine for this short distance.

The background to this question is that we currently cannot get our ISL links to stay up for more than around 30 seconds. The ISL link actually gets created fine(no zonging conflicts), but they then go offline. The process of re-establishing the ISL links then just continues to cycle forever. We currently DON'T have an Extended Fibre License.

We also suspect that the Cisco 1500 may not be correctly configured for fabric mode. i.e. the client was previously using the DWDM purely for IP traffic, and I suspect that one or more settings on the DWDM may have to be tweaked in order to handle fabric. In particular, I believe it's possible to change the "Fixed clocks" rate. Apparently it's meant to be 1062 for fibre.

Sorry for the overly detailed question. Any advice on either the DWDM or Brocade license would be appreciated.

Regards,

Stuart

[harry d brown jr]

Stuart,

Is the license so expensive that it's not worth considering? Get the license and have Cisco help you get it working as part of the deal!

live free or die
harry

[Michael Kerkmann]

Hi Stuart,

as you mentioned the bb_credits is the important fact. You must increase the bb_credits because the sending switch can only send frames as the reciving switch has enove buffers to get this frames. Because of the latence during transmision of the frames, this 16 bb_credits is to small.
And the license politics of Brocade is as it is. If you don't need Extended Fabric you don't pay for it. And if you need it, you pay for it.
It is not microsoft where you get all if you buy Word.

Regards

Michael

[Stuart Walker]

Some responses...

Harry(D Brown Jr), I did neglect to mention that I'm currently working here in China, and it's difficult to get what I'd call, "a reliable response...". I also don't like the idea of paying for something, if it's not necessarily the issue. My understanding of bb_credits has always been that this is the number of 2K packets that can be squeezed onto the link. e.g. if we have a 20km link, then the number of 2K packets is around 10. We currently have 16, so I don't believe that paying for this "Extended Fibre License" will give us a noticable increase in ISL performance. Perhaps someone could convince me otherwise...

Michael(Kerkman), as mentioned above, my understanding of bb_credits is such that I can't see that an increase is particularly required on our 20km link. On longer links, yes - but we are relatively short.

Overall, I really need convincing that this "Extended Fibre License" will actually solve our problem - remembering that we have a basic connectivity problem - not(yet) a performance problem.

Regards,

Stuart

[rakotoarisoa]

Hi,

Here are some indications regarding your question:

* The single mode fiber can support up to 10km.
* So, up to 10 km, the extended fabric license is needed.
* For that distance, you should configure a L1 level to obtain 27 bb-credit for 1GB or 54 bb-credit for 2GB.
* You don't need a link extender for that distance.

Regards,

[Stuart Walker]

rakotoarisoa(and anyone else interested...),

The following is the final solution:

1. We discovered that the clock setting for the cards on the DWDM had in fact been incorrectly set for IP-over-fibre, rather than fabric. Once we changed the clock setting to 1062, our "offline" problem disappeared.

2. As for the Brocade "Extended Fabric License", I would strongly suggest that in our case(~20km), it will mkae very little difference in terms of performance. As previously mentioned, I've always understood the basic theory behind bb_credits is the number of 2K packets that can be squeezed on the link. There is a precise formula for this, but the approximate bb_credits required is half the distance in km's. A few more wouldn't go astray, but I doubt that we would see any performance increase of worth from this license.

Here, perhaps, endeth the story.

Regards,

Stuart Walker