2.4.7. IPSec Crypto HW: IP4 Routed-Forwarding

This section includes summary graphs of VPP Phy-to-Phy packet latency with IPSec encryption used in combination with IPv4 routed-forwarding, with latency measured at 50% of discovered NDR throughput rate. VPP IPSec encryption is accelerated using DPDK cryptodev library driving Intel Quick Assist (QAT) crypto PCIe hardware cards. Latency 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.

Note

Test results have been generated by FD.io test executor jobs csit-vpp-perf-1704-all, with Robot Framework result files csit-vpp-perf-1704-all-<id>.zip archived here.

VPP packet latency in 1t1c setup (1thread, 1core) is presented in the graph below.

Figure 1. VPP 1thread 1core - packet latency for Phy-to-Phy IPSec HW with IPv4 Routed-Forwarding.

CSIT source code for the test cases used for above plots can be found in CSIT git repository:

$ cd $CSIT/tests/perf
$ grep -E "64B-1t1c-.*ipsec.*-ndrdisc" *

40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-aes-gcm-ndrpdrdisc.robot:| tc01-64B-1t1c-ethip4ipsecscale1ip4-ip4base-interfaces-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-aes-gcm-ndrpdrdisc.robot:| tc03-64B-1t1c-ethip4ipsecscale1000ip4-ip4base-interfaces-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-cbc-sha1-ndrpdrdisc.robot:| tc01-64B-1t1c-ethip4ipsecscale1ip4-ip4base-interfaces-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-cbc-sha1-ndrpdrdisc.robot:| tc03-64B-1t1c-ethip4ipsecscale1000ip4-ip4base-interfaces-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-aes-gcm-ndrpdrdisc.robot:| tc01-64B-1t1c-ethip4ipsecscale1ip4-ip4base-tunnels-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-aes-gcm-ndrpdrdisc.robot:| tc03-64B-1t1c-ethip4ipsecscale1000ip4-ip4base-tunnels-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-cbc-sha1-ndrpdrdisc.robot:| tc01-64B-1t1c-ethip4ipsecscale1ip4-ip4base-tunnels-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-cbc-sha1-ndrpdrdisc.robot:| tc03-64B-1t1c-ethip4ipsecscale1000ip4-ip4base-tunnels-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsectptlispgpe-ip4base-cbc-sha1-ndrpdrdisc.robot:| tc01-64B-1t1c-ethip4ipsectptlispgpe-ip4base-cbc-sha1-ndrdisc

VPP packet latency in 2t2c setup (2thread, 2core) is presented in the graph below.

Figure 2. VPP 2threads 2cores - packet latency for Phy-to-Phy IPSec HW with IPv4 Routed-Forwarding.

CSIT source code for the test cases used for above plots can be found in CSIT git repository:

$ cd $CSIT/tests/perf
$ grep -E "64B-2t2c-.*ipsec.*-ndrdisc" *

40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-aes-gcm-ndrpdrdisc.robot:| tc13-64B-2t2c-ethip4ipsecscale1ip4-ip4base-interfaces-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-aes-gcm-ndrpdrdisc.robot:| tc15-64B-2t2c-ethip4ipsecscale1000ip4-ip4base-interfaces-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-cbc-sha1-ndrpdrdisc.robot:| tc13-64B-2t2c-ethip4ipsecscale1ip4-ip4base-interfaces-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-cbc-sha1-ndrpdrdisc.robot:| tc15-64B-2t2c-ethip4ipsecscale1000ip4-ip4base-interfaces-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-aes-gcm-ndrpdrdisc.robot:| tc13-64B-2t2c-ethip4ipsecscale1ip4-ip4base-tunnels-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-aes-gcm-ndrpdrdisc.robot:| tc15-64B-2t2c-ethip4ipsecscale1000ip4-ip4base-tunnels-aes-gcm-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-cbc-sha1-ndrpdrdisc.robot:| tc13-64B-2t2c-ethip4ipsecscale1ip4-ip4base-tunnels-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsecscaleip4-ip4base-tunnels-cbc-sha1-ndrpdrdisc.robot:| tc15-64B-2t2c-ethip4ipsecscale1000ip4-ip4base-tunnels-cbc-sha1-ndrdisc
40ge2p1xl710-ethip4ipsectptlispgpe-ip4base-cbc-sha1-ndrpdrdisc.robot:| tc07-64B-2t2c-ethip4ipsectptlispgpe-ip4base-cbc-sha1-ndrdisc