> I believe we have to take jumbo frames off the table for STP. As I
> understand it (somebody please correct me if I'm wrong!), 802 has
> killed jumbo frames twice, in recent history, once for 1 GigE and once
> for 10 GigE. That basically means that even though some vendors
> support it (Alteon and Extreme, that I know of), it's not going to
> happen.
>
> I have heard this on from two different sources separated by years,
> most recently a few weeks ago from somebody at 3Com who should know.
> He did say that he thought that Alteon was trying to get an IETF
> standardization of jumbo frames, but that seems (to him and me)
> unlikely to go anywhere, because it's not the right place for it to be
> decided.
>
> I guess you could say that running jumbo ethernet is like running
> Myranet or something---an isolated, nonstandard networking
> technology. In other words, SURE, some people do it, but it's
> considered fringe rather than commodity.
>
> I think the upshot is that if STPers focus on jumbo frame performance,
> we're going to look like we're completely out of touch with the real
> world.
>
> So (I'm almost afraid to ask) what's the performance for 1500 byte MTU
> (1K STUs)? The good news is that, you'll probably be able to say < 2%
> CPU, and the bad news is that the data rate is going to be, well,
> lower too.
I totally agree with you on the jumbo frames thing. I remember doing an
extensive analysis of the performance using 1K STUs on alteon's acenic.
I was getting close to 85MB/s - and the bottleneck was the firmware loop
times in the card. There's been some user level I/O work at University
of Cambridge(Ian Pratt) - which is very much related to
STP. They rewrote firmware for the acenic card, and report close to wire
speed with standard ethernet sized frames. Their approach is very
similar to STP, and I'd venture that same results would hold for STP - if
someone rewrote the firmware!
:a
|