[dpdk-dev] Need information on benchmark testing using pktgen

[dpdk-dev] Need information on benchmark testing using pktgen

[abhijit naik]

Hi All,

I am planning to use the pktgen for benchmarking one of the Ethernet
IP. I needed some clarification w.r.t pktgen
Whether throughput, loss and latency can be measured by point to point
connection between test system and DUT(Without switch)? If yes, then
where should pktgen tx and rx should run?
Is support for RFC-2544 test specification complete?
Is their any development activity going on for RFC-2889?

Thank you

-- 
Best wishes,
Abhijit Naik

Re: [dpdk-dev] Need information on benchmark testing using pktgen

[Wiles, Keith]

> On Apr 13, 2020, at 3:55 AM, abhijit naik <abhijitnaik27@gmail.com> wrote:
> 
> Hi All,
> 
> I am planning to use the pktgen for benchmarking one of the Ethernet
> IP. I needed some clarification w.r.t pktgen
> Whether throughput, loss and latency can be measured by point to point
> connection between test system and DUT(Without switch)?

Throughput and packet loss is done within the code, but a percentage of loss is not calculated.
Latency is a bit tricky as the sent from pktgen needs to return to the same port and CPU to get a latency number. The latency must be enabled on the port to get the value. The reason it needs to be the same port/core is the counter used is the per core timestamp register.

> If yes, then
> where should pktgen tx and rx should run?

Not sure about this question, normally pktgen run on one machine targeting the DUT, but you can run pktgen on the machine or the DUT, just a bit more complex of a setup.
> Is support for RFC-2544 test specification complete?

I have a script in the repo that tries to do rfc2544 like test, but may not the best script.
> Is their any development activity going on for RFC-2889?
No.
> 
> Thank you
> 
> -- 
> Best wishes,
> Abhijit Naik