Skip to content

Latest commit

 

History

History
46 lines (32 loc) · 2.98 KB

httpd.md

File metadata and controls

46 lines (32 loc) · 2.98 KB
Raw

Here I we have some PGO benchmarks for httpd.

Test environment

  • Fedora 38
  • Linux kernel 6.4.15
  • AMD Ryzen 9 5900x
  • 48 Gib RAM
  • SSD Samsung 980 Pro 2 Tib
  • Compiler - Clang 16 (16.0.6 from Fedora repositories )
  • httpd version: 2.4.57
  • Other details: Turbo boost is disabled

Benchmark setup

An idea of how to do the benchmark I got from the Rathole benchmark guide for HTTP load. So I implemented the same benchmark for Httpd: Benchmark tool -> httpd (own binary) -> Nginx (system binary).

As a benchmark tool, I use Nighthawk with this command line: taskset -c 15-23 ./nighthawk_client --rps 2000 --duration 120 --connections 2 --concurrency auto --prefetch-connections -v info http://127.0.0.1:8080.

Httpd was tested with this command line: taskset -c 0 httpd -X. httpd.conf content is here: https://gist.github.com/zamazan4ik/cdae9e04e29651e55790ed2b7fab51a0 , all other configuration files are in a default state. I use single-core benchmark setup since I want to load Httpd to 100% on 1 core so I can easily measure the maximum throughput and get the difference in max RPS between Release and PGO builds.

taskset is used everywhere just to reduce OS scheduling noise during the measurements. All measurements are done multiple times (they are almost the same), on the same hardware/software with the same background load (as much as I can guarantee) and the results are compared between runs - they are reproducible with almost no variation between runs.

Optimization steps

Release Httpd is built with CC=clang ./configure --with-included-apr && make command.

Httpd with PGO is built in the following steps:

  • Build Instrumented Httpd with CC=clang CFLAGS="-g -O2 -fprofile-instr-generate=httpd_%p_%m.profraw" LDFLAGS="-fprofile-instr-generate=httpd_%p_%m.profraw" ./configure --with-included-apr && make.
  • Run the instrumented Httpd to collect the profile. As a training workload, I used completely the same workload as used for the benchmark purposes.
  • Prepare the profile for Clang with llvm-profdata merge -output=httpd.profdata httpd_*.profraw
  • Compile Httpd once again with the collected profiles with CC=clang CFLAGS="-g -O2 -fprofile-instr-use=httpd.profdata -Wno-error -Wno-profile-instr-out-of-date" ./configure --with-included-apr && make.

Results

In short, I get the following RPS results from Nighthawk:

  • Release: ~10800 RPS
  • Release + PGO: ~11500 RPS
  • Instrumented (just for reference): ~10400 RPS

More detailed reports from Nighthawk are available here:

According to the tests, PGO helps with optimizing httpd's performance from the throughput perspective at least in the tested scenario.