From 591e09cea0c99467f736e11bd98dc88c037fe811 Mon Sep 17 00:00:00 2001 From: Tibor Frank Date: Fri, 4 May 2018 13:41:43 +0200 Subject: Report: Add TSA plots Change-Id: Ibb8da96518848fef2fc5b0002a84391dea81c573 Signed-off-by: Tibor Frank (cherry picked from commit 4bfd1dcd604dac12d1dd00ba63c0d5e4170a1f2b) --- .../throughput_speedup_multi_core/ipsec.rst | 95 ++++++++++++++++++++++ 1 file changed, 95 insertions(+) create mode 100644 docs/report/vpp_performance_tests/throughput_speedup_multi_core/ipsec.rst (limited to 'docs/report/vpp_performance_tests/throughput_speedup_multi_core/ipsec.rst') diff --git a/docs/report/vpp_performance_tests/throughput_speedup_multi_core/ipsec.rst b/docs/report/vpp_performance_tests/throughput_speedup_multi_core/ipsec.rst new file mode 100644 index 0000000000..678e00fec5 --- /dev/null +++ b/docs/report/vpp_performance_tests/throughput_speedup_multi_core/ipsec.rst @@ -0,0 +1,95 @@ +IPSec Crypto HW: IP4 Routed-Forwarding +====================================== + +Following sections include Throughput Speedup Analysis for VPP multi- +core multi-thread configurations with no Hyper-Threading, specifically +for tested 2t2c (2threads, 2cores) and 4t4c scenarios. 1t1c throughput +results are used as a reference for reported speedup ratio. +VPP IPSec encryption is accelerated using DPDK cryptodev +library driving Intel Quick Assist (QAT) crypto PCIe hardware cards. +Performance is reported for VPP running in multiple configurations of +VPP worker thread(s), a.k.a. VPP data plane thread(s), and their +physical CPU core(s) placement. + +NDR Throughput +-------------- + +VPP NDR 64B packet throughput speedup ratio is presented in the graphs +below for 40ge2p1xl710 network interface card. + +NIC 40ge2p1xl710 +~~~~~~~~~~~~~~~~ + +.. raw:: html + + + +.. raw:: latex + + \begin{figure}[H] + \centering + \graphicspath{{../_build/_static/vpp/}} + \includegraphics[clip, trim=0cm 8cm 5cm 0cm, width=0.70\textwidth]{40ge2p1xl710-64B-ipsechw-tsa-ndrdisc} + \label{fig:40ge2p1xl710-64B-ipsechw-tsa-ndrdisc} + \end{figure} + +CSIT source code for the test cases used for above plots can be found in CSIT +git repository: + +.. only:: html + + .. program-output:: cd ../../../../../ && set +x && cd tests/vpp/perf/crypto && grep -E "64B-(1t1c|2t2c|4t4c)-.*ipsec.*-ndrdisc" * + :shell: + +.. only:: latex + + .. code-block:: bash + + $ cd tests/vpp/perf/crypto + $ grep -E "64B-(1t1c|2t2c|4t4c)-.*ipsec.*-ndrdisc" * + +*Figure 1. Throughput Speedup Analysis - Multi-Core Speedup Ratio - Normalized +NDR Throughput for Phy-to-Phy IPSEC HW.* + +PDR Throughput +-------------- + +VPP PDR 64B packet throughput speedup ratio is presented in the graphs +below for 40ge2p1xl710 network interface card. + +NIC 40ge2p1xl710 +~~~~~~~~~~~~~~~~ + +VPP PDR 64B packet throughput in 1t1c setup (1thread, 1core) is presented +in the graph below. PDR measured for 0.5% packet loss ratio. + +.. raw:: html + + + +.. raw:: latex + + \begin{figure}[H] + \centering + \graphicspath{{../_build/_static/vpp/}} + \includegraphics[clip, trim=0cm 8cm 5cm 0cm, width=0.70\textwidth]{40ge2p1xl710-64B-ipsechw-tsa-pdrdisc} + \label{fig:40ge2p1xl710-64B-ipsechw-tsa-pdrdisc} + \end{figure} + +CSIT source code for the test cases used for above plots can be found in CSIT +git repository: + +.. only:: html + + .. program-output:: cd ../../../../../ && set +x && cd tests/vpp/perf/crypto && grep -E "64B-(1t1c|2t2c|4t4c)-.*ipsec.*-pdrdisc" * + :shell: + +.. only:: latex + + .. code-block:: bash + + $ cd tests/vpp/perf/crypto + $ grep -E "64B-(1t1c|2t2c|4t4c)-.*ipsec.*-pdrdisc" * + +*Figure 2. Throughput Speedup Analysis - Multi-Core Speedup Ratio - Normalized +PDR Throughput for Phy-to-Phy IPSEC HW.* -- cgit 1.2.3-korg