Re: MSA’s and network architecture

Saku Ytti saku at ytti.fi
Wed May 18 08:46:45 UTC 2022


On Wed, 18 May 2022 at 11:35, Mark Tinka <mark at tinka.africa> wrote:

> Unless you are truly desperate and/or happy to get stuck in vendor-land,
> always wise to be slightly behind the curve when it comes to optics.

Agreed, if possible do boring things and get boring results.

Even in vendor land, a boring result is not guaranteed. Vendor had one
100GE LR4 SKU approved for their platform and it had problems
performing under some conditions. Issue was with something LBW (loop
bandwidth?) some optics have this at 10MHz others 20MHz, and the
former does not perform with certain jitter, latter does and the
former was only one approved by the vendor. Vendor addressed this by
removing approval for the only version that was approved and approved
another one. It was not blatantly broken, it would almost certainly
work fine in your lab, as environmental conditions needed to apply.

Also testing goes only so far, because vendors change suppliers and
hardware all the time, without changing SKU. Cisco is the only vendor
I know who has very detailed change logs of what they are doing to the
SKUs with each change. Single order might contain multiple versions of
the SKU and the versions may not behave the same. We recently had this
problem with 3rd party optic, where a new version of SKU didn't work
for us, and we had to discover it in the field and we only knew of the
SKU change after the problem occurred. Of course probably more than 99
SKU changes out of 100 are invisible to end users.




--
  ++ytti


More information about the NANOG mailing list