[j-nsp] TCP

[Payam Chychi]

Hi Johan,

This sounds like a network issue, im actually dealing with the same 
thing with one of my mpls providers.

Latency does of course play a factor however, latency is a bidirectional 
influence and not asymmetric. No reason as to why you should be getting 
400 one way and 60 the other.
Only thing comes to mind is different paths the packets may travel as 
most load balancing is of course done based on a src/dst layer3 hash so 
there could be a switch fabric or a path somewhere that is causing the 
congestion.

You should do what im doing, grant your provider a maintenance window to 
take down your circuit and do an end to end throughput test and make 
sure they provide you with the stats.
If they can get 1:1 capacity then look at your optics, interfaces, any 
bundled links at both.

Cheers
Payam



On 2014-11-19, 1:18 PM, Johan Borch wrote:
> Hi!
>
>
> I'm doing some performance troubleshooting between two linux systems, the
> servers are located in each end of an L3VPN, with a bunch of routers
> between them.
>
> Using Iperf and UDP I get ~1Gbps in both directions
> Using iperf and TCP i get ~400Mbps in one direction and ~60Mbps in the
> other direction
>
> Could this still be a network problem or should I dig on the linux side?
>
> Johan
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp