mirror of
https://github.com/ossrs/srs.git
synced 2026-10-28 22:59:37 +03:00
SRS gets performance tools that load and profile SRS in Docker for every protocol: a lean load tool, an internal skill that runs an improvement round, and a measurement method that holds up to a 10% verdict. The same tools then measure the Go proxy. The first runs found one SRS fix and three Go proxy fixes. The load tool, a standalone Go module in cmd/perf (bin/srs-perf): - Per-core worker processes shard the clients and report counters each second. Lean clients for every protocol: RTMP publish and play, HTTP-FLV, HTTP-TS, HLS, DASH, RTSP, SRT, WHIP and WHEP. No decode; frames and lag come from timestamps. - 17 scenarios: each protocol's fan-out, publishers (rtmp-pub, whip-pub), many-streams, cross-protocol bridges (srt-rtmp, srt-whep, whip-rtmp, whip-flv), and delay probes for the RTMP pause, forward and edge. - run ramps by doubling to a capacity, then holds at 50% and 90% of it. compare checks that the setups match and reports capacity and CPU per client at the 10% line. - make perf-image builds an optimized SRS with symbols and frame pointers, with perf, FlameGraph and FFmpeg. SRS runs on one pinned core and is configured only by env files, one per scenario. The measurement method: - A flat CBR source, checked to ±3% per second, since SRS CPU follows the bitrate. - SRS CPU from its cgroup once a second. Holds have no profiler in the measuring window; each count gets its own separate profile hold. - perf-ab.sh builds two revisions and runs them side by side on separate cpusets. compare gives the 95% CI of the paired per-second difference. - perf-verify.sh checks every run: capacity, healthy steps, profiles and symbols, noise, and what bound the run. The internal-perf-for-srs skill: - SKILL.md holds the rules, the shared tools and the round workflow. A Target Router loads references/cpp-server.md for the C++ media server or references/go-server.md for the Go proxy. - Each reference covers its tools, topology, scenarios, measuring, how to read its profiles (Linux perf, or Go pprof and trace), A/B, the verified scenarios with their capacity and top bottleneck, and how to start a round. - srs-develop routes performance, benchmark, load test and profile tasks to it. perf-skill-test.sh checks the text against the scenario registry. - Reports are git-ignored, since their numbers hold only for the machine that made them. - The skills now name each server by its job: the C++ media server and the Go proxy, with no "generation". C++ media server, one core, Linux arm64 in Docker: - Fix: an empty SrsSharedPtr holds no reference count, so each fan-out copy of a message does one malloc instead of three. Covered by srs_utest_ai05.cpp. - Capacity: rtmp-rtmp, rtmp-flv, rtmp-hls, whip-rtmp and whip-flv 12,800; rtmp-dash and srt-rtmp 6,400; rtmp-ts and rtmp-rtsp 3,200; whip-whep 2,000 (cap); rtmp-whep, srt-whep, srt-srt and rtmp-pub 1,280; whip-pub 1,000 (cap); many-streams 640 streams. - Top bottlenecks found, not fixed: the RTMP fan-out copy, _st_epoll_dispatch with EAGAIN reads, 4 KB reads and writes for HLS and DASH segments, one write per message for HTTP-TS and two per packet for RTSP, the SRTP replay window in libsrtp, and libsrt for SRT. - Delays: the paused RTMP player's 500 ms sleep, and the forward's and edge publish's 500 ms receive timeout, each adding about 220 ms of lag. The Go proxy, cmd/proxy, on one core in front of 1 to 4 SRS origins: - perf-up.sh -target proxy, run -target proxy, which samples the proxy and each origin, Go profiles in each profile hold (CPU, allocations, mutex, block, trace, GC), and proxy A/B. - Fix: HTTP-FLV, HTTP-TS and HLS relay the origin's error status, headers and body, rather than a 500 with the origin's address. - Fix: a WebRTC STUN binding routes by its ufrag, not a cached client address. WHIP and WHEP failed from 160 clients on three origins. - Capacity on three origins: rtmp-flv and rtmp-hls 12,800; rtmp-rtmp and rtmp-ts 6,400; whip-whep 2,000 (cap); srt-srt and rtmp-pub 1,280; whip-pub 1,000 (cap); many-streams 1,280 streams. --------- Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
221 lines
7.2 KiB
Go
221 lines
7.2 KiB
Go
package main
|
|
|
|
import (
|
|
"encoding/json"
|
|
"fmt"
|
|
"io"
|
|
"sync"
|
|
"time"
|
|
)
|
|
|
|
// runSmoke publishes the source with the scenario's publisher and plays it
|
|
// with 100 players of the scenario's protocol for 10s, or for three segment
|
|
// durations, by the scenario's gap and lag rules.
|
|
func runSmoke(output io.Writer, cmd command) error {
|
|
const clients = 100
|
|
s := scenarios[cmd.scenario]
|
|
if s.Unit == "probes" {
|
|
return runDelaySmoke(output, cmd)
|
|
}
|
|
if s.Unit != "clients" {
|
|
return runPublisherSmoke(output, cmd, clients)
|
|
}
|
|
dial, err := playerDialer(s.Player)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
tags, flatness, err := loadSource(cmd.media)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
// SRS, or the proxy with its origins beside it, is pid 1 of the load
|
|
// container's pid namespace.
|
|
measured, origins, err := targetSetup(withSource(setup{}, cmd.media, flatness), cmd.target, proxyAPI, "/proc", 1)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
// One stream on SRS, or one on each origin behind the proxy, with the
|
|
// players split over them.
|
|
apis := originAPIs(origins)
|
|
published, err := startStepStreams(scenarios[cmd.scenario].Publisher, "perf-smoke", cmd.media, tags, apis)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
stop := make(chan struct{})
|
|
var stopOnce sync.Once
|
|
var stopErr error
|
|
stopAll := func() { stopOnce.Do(func() { close(stop); stopErr = stopStreams(published) }) }
|
|
// A publisher that does not stop must not hold up an error's return.
|
|
defer func() { go stopAll() }()
|
|
if measured.Shares, err = targetShares(s, apis, published, "perf-smoke", clients); err != nil {
|
|
return err
|
|
}
|
|
names := streamNames(published)
|
|
timeline, err := lagTimeline(cmd.media, tags)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
players := newWorkerPlayers(clients, timeline)
|
|
players.gap = s.gap()
|
|
players.streams = playerStreamIndex(0, clients, len(names))
|
|
errors := make(chan error, clients)
|
|
var wg sync.WaitGroup
|
|
for i := 0; i < clients; i++ {
|
|
wg.Add(1)
|
|
go func(id int) {
|
|
defer wg.Done()
|
|
c, err := dial()
|
|
if err != nil {
|
|
errors <- fmt.Errorf("player %d dial: %w", id, err)
|
|
return
|
|
}
|
|
report := make(chan rtmpMessage, 128)
|
|
go func() {
|
|
for m := range report {
|
|
players.received(id, m)
|
|
}
|
|
}()
|
|
if err := c.play(names[players.streams[id]], stop, report); err != nil {
|
|
select {
|
|
case <-stop:
|
|
default:
|
|
errors <- fmt.Errorf("player %d: %w", id, err)
|
|
}
|
|
}
|
|
close(report)
|
|
}(i)
|
|
}
|
|
// SRS shares the load container's pid namespace as pid 1.
|
|
ticks, err := clockTicks()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
sampler := newProcSampler("/proc", "/sys/fs/cgroup", 1, nil, ticks).withOrigins(origins)
|
|
if _, err := sampler.sampleAt(time.Now()); err != nil {
|
|
return err
|
|
}
|
|
ticker := time.NewTicker(time.Second)
|
|
defer ticker.Stop()
|
|
var samples []sample
|
|
var maxLag int64
|
|
for len(samples) < s.smokeSeconds() {
|
|
select {
|
|
case err := <-errors:
|
|
return err
|
|
case now := <-ticker.C:
|
|
counters := players.sample(streamTimestamps(published), false)
|
|
usage, err := sampler.sampleAt(now)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
second := sample{Second: len(samples) + 1, Frames: counters.Frames, Segments: counters.Segments, Bytes: counters.Bytes, LagMillis: counters.LagMillis}
|
|
second.SRSCPUPercent, second.SRSCPUUsageUsec, second.SRSRSSBytes = usage.srsCPUPercent, usage.srsCPUUsageUsec, usage.srsRSSBytes
|
|
second.OriginCPUPercent, second.OriginCPUUsageUsec = usage.originCPUPercent, usage.originCPUUsageUsec
|
|
samples = append(samples, second)
|
|
if counters.LagMillis > maxLag {
|
|
maxLag = counters.LagMillis
|
|
}
|
|
}
|
|
}
|
|
for i := range players.last {
|
|
at := players.last[i].Load()
|
|
if at == 0 || time.Since(time.Unix(0, at)) > s.gap() {
|
|
return fmt.Errorf("player %d has no recent frame", i)
|
|
}
|
|
}
|
|
if limit := s.lagLimit(); limit > 0 && maxLag > limit {
|
|
return fmt.Errorf("player lag %dms exceeds %dms", maxLag, limit)
|
|
}
|
|
stopped := make(chan struct{})
|
|
go func() { stopAll(); close(stopped) }()
|
|
select {
|
|
case <-stopped:
|
|
if stopErr != nil {
|
|
return stopErr
|
|
}
|
|
case <-time.After(3 * time.Second):
|
|
return fmt.Errorf("publisher did not stop")
|
|
}
|
|
done := make(chan struct{})
|
|
go func() { wg.Wait(); close(done) }()
|
|
select {
|
|
case <-time.After(3 * time.Second):
|
|
return fmt.Errorf("players did not stop")
|
|
case <-done:
|
|
}
|
|
r := result{Setup: measured, Scenario: scenarios[cmd.scenario], Steps: []step{{Mode: "smoke", Clients: clients, SettleSeconds: 2, Samples: samples, Health: health{Healthy: true}, Validity: validity{Valid: true}}}, Health: health{Healthy: true}, Validity: validity{Valid: true}}
|
|
return json.NewEncoder(output).Encode(summarizeResult(r, "smoke"))
|
|
}
|
|
|
|
// runPublisherSmoke runs a publisher or many-streams scenario's clients in
|
|
// this process, as a worker runs them, for 10s: every publisher keeps its
|
|
// pace, every player gets frames within the lag limit, and none fails, the
|
|
// last second judging a client that never sent or got anything.
|
|
func runPublisherSmoke(output io.Writer, cmd command, clients int) error {
|
|
s := scenarios[cmd.scenario]
|
|
tags, flatness, err := loadSource(cmd.media)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
measured, origins, err := targetSetup(withSource(setup{}, cmd.media, flatness), cmd.target, proxyAPI, "/proc", 1)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
timeline, err := lagTimeline(cmd.media, tags)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
runtime, err := startClients(s, clients, []string{"perf-smoke"}, 0, cmd.media, tags, timeline)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
closed := false
|
|
defer func() {
|
|
if !closed {
|
|
runtime.close()
|
|
}
|
|
}()
|
|
ticks, err := clockTicks()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
sampler := newProcSampler("/proc", "/sys/fs/cgroup", 1, nil, ticks).withOrigins(origins)
|
|
if _, err := sampler.sampleAt(time.Now()); err != nil {
|
|
return err
|
|
}
|
|
ticker := time.NewTicker(time.Second)
|
|
defer ticker.Stop()
|
|
var samples []sample
|
|
for len(samples) < s.smokeSeconds() {
|
|
now := <-ticker.C
|
|
counters := runtime.sample(nil, len(samples)+1 == s.smokeSeconds())
|
|
usage, err := sampler.sampleAt(now)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
second := sample{Second: len(samples) + 1, Frames: counters.Frames, Bytes: counters.Bytes, LagMillis: counters.LagMillis, Failures: counters.Failures}
|
|
second.SRSCPUPercent, second.SRSCPUUsageUsec, second.SRSRSSBytes = usage.srsCPUPercent, usage.srsCPUUsageUsec, usage.srsRSSBytes
|
|
second.OriginCPUPercent, second.OriginCPUUsageUsec = usage.originCPUPercent, usage.originCPUUsageUsec
|
|
samples = append(samples, second)
|
|
if counters.Failures > 0 {
|
|
return fmt.Errorf("second %d: %d %s failed", second.Second, counters.Failures, s.Unit)
|
|
}
|
|
if limit := s.lagLimit(); limit > 0 && counters.LagMillis > limit {
|
|
return fmt.Errorf("second %d: lag %dms exceeds %dms", second.Second, counters.LagMillis, limit)
|
|
}
|
|
}
|
|
if measured.Shares, err = targetShares(s, originAPIs(origins), nil, "perf-smoke", clients); err != nil {
|
|
return err
|
|
}
|
|
done := make(chan struct{})
|
|
go func() { runtime.close(); close(done) }()
|
|
closed = true
|
|
select {
|
|
case <-time.After(3 * time.Second):
|
|
return fmt.Errorf("%s did not stop", s.Unit)
|
|
case <-done:
|
|
}
|
|
r := result{Setup: measured, Scenario: s, Steps: []step{{Mode: "smoke", Clients: clients, SettleSeconds: 2, Samples: samples, Health: health{Healthy: true}, Validity: validity{Valid: true}}}, Health: health{Healthy: true}, Validity: validity{Valid: true}}
|
|
return json.NewEncoder(output).Encode(summarizeResult(r, "smoke"))
|
|
}
|