Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [HSSG] 40G MAC Rate Discussion



On Tue, 3 Apr 2007, Shimon Muller wrote:

> Nothing magic about 40Gb for servers, just as there is nothing magic 
> about a 10x scaling factor, except for tradition. 40Gb happens to be in 
> the right ballpark for server needs in that timeframe, and it will be at 
> the right cost.

I represent a customer view here.

My coworkers who run servers seem to have no problem with link aggregation 
and they have a history with four port FE cards and they're now doing LAG 
with gig ports when they need it. The only 10GE attached servers we have 
exist in the network lab for cheap packet generation.

Myself, being a network engineer, I see operational problems with more 
than 4 links in a LAG but 4 is quite ok. I therefore see little reason for 
a 40G server speed, as this can be solved by 4*10GE LAG. A 10x scaling 
factor makes sense because 10 links in a LAG starts to create operational 
problems with proper load balancing, how you manage them both physically 
and logically, and a host of other problems that is better solved on the 
hardware layer.

I see more reason for a 40G optical standard that basically extends the 
OC192 10GE WAN PHY standard to OC768 as this would enable ethernet to be 
carried in current state of the art DWDM systems. This would of course be 
a moot point if we think that most DWDM vendors will have a 100GE DWDM 
transponder by the time the 100GE standard is ready, and a 40GE OC768 
compatible standard would take the same amount of time to implement as the 
100GE variant.

I do think that it should be taken into account that a 100GE standard will 
be carried over DWDM, and mechanisms should be implemented for 
fast-failure-detection, mostly of the "there is a one-way fiber 
break"-variant and the "the BER is high so the link is not good". This is 
perhaps something that should be discussed in another thread...

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se