ISR 4321 throughput max

[iverson.justin]

Question on ISR 4321, right now our device is setup for 100000 kb/s throughput, however when testing iperf on our private WAN connection (which is 100mb/s) we are only getting 42mbps. We are likely hitting a throughput limit on this device?  Topology is 

Server(iperf)-->switch-->ISR-->WAN MPLS--->ASR-->Switch-->Server (iperf)

[Joseph W. Doherty]

100000 kb/s throughput - Do you mean you have 100 Mbps performance license?

Same performance, both directions?

Are you testing throughput in just one direction, or both directions, at the same time?

For iPerf, are you bandwidth testing using TCP or UDP?

WAN handoffs are FE or gig?

IPerf reporting any lost packets?

[iverson.justin]

WAN handoff is Gig, using iperf tcp and yes I used the -R switch to check iperf in reverse direction. I am not sure if iperf is dual or single stream.  Pretty sure its single stream.  Either transmit or receiving I am only getting 42mbps

[Joseph W. Doherty]

To insure I understand, your WAN is a gig physical hand-off with a logical 100 Mbps cap?

[iverson.justin]

Correct gig ethernet handoff and 100mbps through provider cap.  

[Joseph W. Doherty]

Then it's possible, gig bursts are being dropped by your provider, which will slow TCP transmission rate.

That noted, I wouldn't expect such an impact to drive your rate to 42 Mbps; further, the 4321 should enforce its cap by shaping (BTW, the 4321 cap is applied to all traffic flowing through the router, so when doing these tests, check the aggregate throughput of all interfaces, ingress or egress (but not both, together).

Also try (if iPerf supports), unidirectional bandwidth tests, using UDP, at different "fixed" data rates, including over 100 Mbps, and "see" what rate you get on the far side.

[iverson.justin · Accepted solution]

Boost performance license resolved this issue on 4321

Index 11 Feature: booster_performance

[Joseph W. Doherty]

Interesting!  Although, still leaves the "mystery" why you didn't get your original licensed throughput.

Thanks for the follow-up.