07-31-2013 08:54 AM
We have several CER routers which are displaying throughput issues when using SSH and SCP. These are on completely un-rate limited connections and duplicated in a lab environment where nothing but two laptops are running through these CER routers.
All are boot/monitor/PBIF/firmware upgraded to the latest versions. Initially we thought it might be one of our fiber providers, but again we were able to duplicate this in a lab environment.
We are going to try downgrading firmware in steps to see if we can replicate, or eliminate the issue - however we do need 5.4 or better due to running 10G fiber. We have many customers behind this router suffering with SSH oddities, and for developers and admins this is a big problem, so it is a big problem for us. Has anyone seen or experienced similar? We also have NO QoS in place as we do not shape customer traffic other than a raw rate limit based on subscription, in the lab though we are running completely un-rate limited "wire speed" tests.
Thanks for any thoughts!
08-02-2013 09:04 PM
As a follow up, we have replaced the CER-2000RT and we thought that everything was fixed... until we re-enabled BGP. Memory, cam tables, etc etc are not full so versus the paper specification we are not pushing this router to its limits but SSH/SCP throughput is still limited over copper or fiber. On two high powered servers on gigabit eth ports coming off CER-2000RTs fully updated (5.5, pbif etc) I get ~24MB/s peak throughput between the devices.
We also have 2x 10gbit fiber links in a couple of these, but they are not being pushed to their limits throughput wise nor is PPS anywhere near maximum throughput. I think during my test we totalled 400k pps on one of the routers.
08-08-2013 11:38 AM
Ive done a little checking into this and unfortunately havent found much as to why you may be seeing this. Possibly related to the defect below from the 5.50c Release notes. Have you considered opening a TAC case so that they can further diagnose your configuration?