[j-nsp] Juniper M20 LACP oddness.

Richard A Steenbergen ras at e-gerbil.net
Fri Apr 24 18:17:42 EDT 2009


On Fri, Apr 24, 2009 at 03:11:04PM -0600, Jonathan Call wrote:
> 
> Why do you believe this is a hash settings issue? The egress load
> seems evenly distributed between the two active interfaces. When the
> ae0 interface plateaus at 1Gbps , both physical interfaces plateau at
> ~500Mbps.  I've also read that layer-4 hashing can be CPU/memory
> intensive. Would just layer-3 be sufficient?

If hashing takes any memory you're doing it wrong, this is stateless
operation with no memory impact and no CPU impact since its done in
hardware.

> The other end is a Cisco, but since this issue is with egress traffic
> I don't really think that is important, unless I do change the
> load-balancing settings.

Are you able to get 1Gbps on any one individual link, if you bring it 
down to only 1 active ae member? And, just to get the obvious question 
out of the way, you aren't congested on the egress port from the Cisco 
are you? :)

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)


More information about the juniper-nsp mailing list