[c-nsp] XENPAK packet loss

[Mikael Abrahamsson]

On Mon, 15 Jun 2009, FF wrote:

> On a longer span (140km) link, I have the same setup, with two EDFAs. 
> The input power comes in right around -24dbm. It tests clean mostly, but 
> drops 2 packets out of 10,000 when empty. More traffic, more drops.  I 
> figured it was a strength problem (-24dbm is right at the edge). So went 
> in with a some fiber cleaners and rejumpered everything. Got it down to 
> -22dbm which should be well within the tolerances. If anything the 
> problem got worse, not better.

Is this link dispersion compensated? Otherwise that is most likely your 
problem, 1GE rarely get chromatic dispersion (CD) problems, 10GE much more 
so.

> So now I have two spans, of drastically different lengths, both dropping 
> packets directly related to the amount of traffic moving over them. Only 
> BGP and static routes are on these. All are 6500/SUP32 or SUP720s 
> running 12.22-33SXH5 or SXHI1.

Which indicates a constant BER (bit error rate) which is consistant with 
CD induced BER.

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