diff options
Diffstat (limited to 'doc/trex_analytics.asciidoc')
| -rwxr-xr-x | doc/trex_analytics.asciidoc | 18 |
1 files changed, 17 insertions, 1 deletions
diff --git a/doc/trex_analytics.asciidoc b/doc/trex_analytics.asciidoc index 6635a068..a5b223ea 100755 --- a/doc/trex_analytics.asciidoc +++ b/doc/trex_analytics.asciidoc @@ -23,6 +23,22 @@ endif::backend-xhtml11[] = TRex performance trend +This document tracks the performance of TRex Stateless use-cases of various hardware and NICS. +The intention is to track the TRex software performance and not the NIC driver performance, however from this document you can learn the difference between different driver implementation (cycles/packets) +Performance is measured in a metric called MPPS/Core (Norm), which stands for Mega-Packet per second (Norm). + + +How the calculation is made? + + + +We measure the MPPS of the DUT in its best working point. + +The measurement is then scaled to fit 100% CPU utilization, and finally divided by the number of cores the device has, +to get a normalized MPPS/Core which is denoted as MPPS/Core (Norm) + + + +How to run these tests? + +[source,python] +---- + trex-core/scripts> ./run_regression --stl -t <name of test> --cfg setups/<name of setup> +---- + == Setup: XL710 (4x40GbE) trex-08 .Setup Details @@ -132,7 +148,7 @@ include::build/images/kiwi02_trend_stats.csv[] |================= | Server | Cisco UCS 240M3 | CPU: | 2 sockets x Intel(R) Xeon(R) CPU E5-2667 v3 @ 3.20GHz, 8 cores -| NICs: | 2 NICS x ( 2x 40GbE ) Intel X710, Total 8 ports +| NICs: | 1 NIC x ( 2x 40GbE ) Cisco 1300 VIC | NUMA | Unknown due to ESXi | Memory | 2x16GB | PCIe | 2x PCIe16 + 2xPCIe4 |