[c-nsp] FW: Overruns

[Nick Hilliard]

On 27/01/2011 16:00, Brett Frankenberger wrote:
> Of course he is.

It certainly adds extra resilience if you have an individual bearer link 
failure.  But it also not unreasonable to assume that adding links into a 
LAG will increase its throughput.  Unfortunately, this is often not the 
case on lots of Cisco hardware - including sup720 based equipment.

Taking your example below, if you had an 8 port LAG and one of them failed, 
this would drop you down to a 4 port LAG in terms of throughput.  If you 
aren't expecting it, this is astonishing behaviour and have pretty serious 
consequences.

This wouldn't be so much of a problem if the restriction were well 
documented by Cisco.  Unfortunately, it's not - you have to dig to find it, 
and then infer that the same restrictions in CatOS also apply to IOS.  I 
guess it's not something that Cisco like to make too much noise about.

Nick


> With five links, assuming the traffic hashes evenly
> across the 8 buckets, he effectvly has 4GBps of throughput available.
> If one of the five links fails, he still has 4GBps of throughput
> available.
>
> With four links, assuming the traffic hashes evenly across the 8
> buckets, he effectively has 4Gbps of throughput available.  But if one
> of the four links fails, then he'll hash at 3:3:2 and effectively have
> 2.67Gbps available.
>
> In other words, the fifth link doesn't add any throughput benefit in
> the "everything is working" case -- four or five active links offers
> the same throughput -- but it offers a significant redundancy benefit.
> With five links, loss of one link is no impact to cpacity; with four,
> loss of one link is a 33% reduction in capacity.
>
>       -- Brett