commit 68ca4f24d6502e6e3e6eeead4ba5a2cb5580736d
author Jonas Oreland <jonaso@webrtc.org> Tue Dec 09 21:03:24 2025
committer WebRTC LUCI CQ <webrtc-scoped@luci-project-accounts.iam.gserviceaccount.com> Tue Dec 09 23:33:33 2025
tree e7f76f6bfa8f916633ff511b57d44a9883455cdd
parent 62b6eb25299dee6af2990399ae8c2db3353191e2

Create a simulated time benchmark (with packet loss) for DtlsInStun and PQC

The submitted version is *not* run automatically, only when specifying
--bench_iterations=100.

It is to be considered a development/debugging tool.

It currently tests these configurations,
which could be a tentative release plan.

a) DTLS1.2
b) DTLS1.3
c) DTLS1.3 + PQC + DTLS-IN-STUN

The test runs each of this configuration in a set of different
environments, e.g packet loss, server candidates etc and numbers
look great, if rolled out like above, i.e. launch dtls-in-stun
and the same time as PQC.

sh> out/Release/rtc_p2p_unittests --gtest_filter='*Bench*/*' --bench_iterations=25 --bench_server_candidates=1  2>&1 | grep RES
(dtls_ice_integrationtest.cc:942): loss: 0 server_interface_count: 1 [ client: [ dtls: client/1.2 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.2 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 850 p50: 850 avg: 850 p95: 850
(dtls_ice_integrationtest.cc:942): loss: 0 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.3 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 750 p50: 750 avg: 750 p95: 750
(dtls_ice_integrationtest.cc:942): loss: 0 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 1 dtls_in_stun: 1  ][ server: [ dtls: server/1.3 ice: controlled pqc: 1 dtls_in_stun: 1  ]] RESULT: p10: 650 p50: 650 avg: 650 p95: 650

(dtls_ice_integrationtest.cc:942): loss: 5 server_interface_count: 1 [ client: [ dtls: client/1.2 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.2 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 850 p50: 850 avg: 940 p95: 1300
(dtls_ice_integrationtest.cc:942): loss: 5 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.3 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 750 p50: 750 avg: 838 p95: 1150
(dtls_ice_integrationtest.cc:942): loss: 5 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 1 dtls_in_stun: 1  ][ server: [ dtls: server/1.3 ice: controlled pqc: 1 dtls_in_stun: 1  ]] RESULT: p10: 650 p50: 650 avg: 666 p95: 700

(dtls_ice_integrationtest.cc:942): loss: 10 server_interface_count: 1 [ client: [ dtls: client/1.2 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.2 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 850 p50: 900 avg: 1106 p95: 2060
(dtls_ice_integrationtest.cc:942): loss: 10 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.3 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 750 p50: 800 avg: 910 p95: 1200
(dtls_ice_integrationtest.cc:942): loss: 10 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 1 dtls_in_stun: 1  ][ server: [ dtls: server/1.3 ice: controlled pqc: 1 dtls_in_stun: 1  ]] RESULT: p10: 650 p50: 650 avg: 702 p95: 840

(dtls_ice_integrationtest.cc:942): loss: 25 server_interface_count: 1 [ client: [ dtls: client/1.2 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.2 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 870 p50: 1300 avg: 6208 p95: 11820
(dtls_ice_integrationtest.cc:942): loss: 25 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.3 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 750 p50: 1300 avg: 1358 p95: 2420
(dtls_ice_integrationtest.cc:942): loss: 25 server_interface_count: 1 [ client: [ dtls: client/1.3 ice: controlling pqc: 1 dtls_in_stun: 1  ][ server: [ dtls: server/1.3 ice: controlled pqc: 1 dtls_in_stun: 1  ]] RESULT: p10: 650 p50: 750 avg: 942 p95: 1800

I.e. dtls-in-stun+pqc is faster then the other combinations even with
packet loss.

However, the test also shows that we still have bugs in the code!
sh> out/Release/rtc_p2p_unittests --gtest_filter='*Bench*/*' --bench_iterations=25 --bench_pct_loss=25 --bench_server_candidates=2  2>&1 | grep RES
(dtls_ice_integrationtest.cc:942): loss: 25 server_interface_count: 2 [ client: [ dtls: client/1.2 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.2 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 870 p50: 1300 avg: 2488 p95: 8480
(dtls_ice_integrationtest.cc:942): loss: 25 server_interface_count: 2 [ client: [ dtls: client/1.3 ice: controlling pqc: 0 dtls_in_stun: 0  ][ server: [ dtls: server/1.3 ice: controlled pqc: 0 dtls_in_stun: 0  ]] RESULT: p10: 750 p50: 1200 avg: 1858 p95: 7600
(dtls_ice_integrationtest.cc:942): loss: 25 server_interface_count: 2 [ client: [ dtls: client/1.3 ice: controlling pqc: 1 dtls_in_stun: 1  ][ server: [ dtls: server/1.3 ice: controlled pqc: 1 dtls_in_stun: 1  ]] RESULT: p10: 650 p50: 950 avg: 6.91307e+06 p95: 6.91213e+07

Notice outlier!

BUG=webrtc:367395350

Change-Id: I44c0e1df33108d0078f1bf88804f48e1ce469b88
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/432180
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#46439}