[c-nsp] GSR12008|GRP-B|4OC12/ATM-MM-SC|3GE-GBIC-SC throughput?

[Jason Lixfeld]

On 15-Apr-09, at 1:30 PM, Peter Rathlev wrote:

> This may or may not be relevant, but depending on how much extra  
> latency
> the 12008 introduces you might still have a client side limitation  
> doing
> TCP. Reasonable TCP window sizes and effective sender side congestion
> control are needed.

Good advice.  Didn't consider that.

> How much latency end-to-end in the setup with/without the 12008?

12ms (Toronto to New York and back) with the 12008.  Haven't hair  
pinned a port on the New York Fore yet, so can't determine latency  
without the 12008.

> Making IPerf use a defined load in a UDP stream (e.g. 100 Mbps) and  
> then
> measuring the loss would overcome this limitation of course.

I'd happily test that, however we don't have a machine on-site in New  
York, unless you (or anyone else, for that matter ;)) happen to have a  
box hanging off NYIIX that I could dump traffic to over our NYIIX link.

> Regards,
> Peter
>
>
>