Add shenandoah-jdk28: generational vs SATB Shenandoah, G1 and ZGC on a Spring Boot service (JEP 535 status probe included)
Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01HmvoHH3ZnCEaPNXRpDxDZe
This commit is contained in:
Claude
@@ -72,6 +72,7 @@ the part most "ZGC is faster" posts leave out.
|
|||||||
| [`analysis/GcLogParser.java`](analysis/src/com/ankurm/zgc/analysis/GcLogParser.java) | Unified-log parser that keeps pauses and concurrent phases apart | none |
|
| [`analysis/GcLogParser.java`](analysis/src/com/ankurm/zgc/analysis/GcLogParser.java) | Unified-log parser that keeps pauses and concurrent phases apart | none |
|
||||||
| [`jmh/`](jmh/src/main/java/com/ankurm/zgc/jmh) | Throughput and barrier microbenchmarks | JMH 1.37 |
|
| [`jmh/`](jmh/src/main/java/com/ankurm/zgc/jmh) | Throughput and barrier microbenchmarks | JMH 1.37 |
|
||||||
| [`jdk27-memory/`](jdk27-memory) | Spring Boot 4.1 service measured on JDK 25, 26 and 27: compact object headers (JEP 534) and G1 everywhere (JEP 523), heap footprint, container sizing with real cgroup limits. Its README and `output/` replace a `docs/` folder; the write-up is [the post](https://ankurm.com/jdk-27-compact-object-headers-g1-default-memory-spring-boot/) | Spring Boot 4.1.1, three JDKs |
|
| [`jdk27-memory/`](jdk27-memory) | Spring Boot 4.1 service measured on JDK 25, 26 and 27: compact object headers (JEP 534) and G1 everywhere (JEP 523), heap footprint, container sizing with real cgroup limits. Its README and `output/` replace a `docs/` folder; the write-up is [the post](https://ankurm.com/jdk-27-compact-object-headers-g1-default-memory-spring-boot/) | Spring Boot 4.1.1, three JDKs |
|
||||||
|
| [`shenandoah-jdk28/`](shenandoah-jdk28) | Spring Boot 4.1 service on G1, generational ZGC, SATB Shenandoah and generational Shenandoah (JEP 535, JDK 28 EA): latency under fixed-rate load, GC pauses, throughput, RSS. Its README and `output/` replace a `docs/` folder | Spring Boot 4.1.2, Temurin 25 and 28 EA |
|
||||||
| [`flags/`](flags) | The flag-by-flag production cheat sheet: `MaxRAMPercentage` vs a hardcoded `-Xmx` (one fails safely, one gets the container SIGKILLed), `-XX:-UseContainerSupport`, the `ActiveProcessorCount` lie, and which diagnostic flags (`HeapDumpOnOutOfMemoryError`, `NativeMemoryTracking`) are worth leaving on in prod — on real Docker containers, JDK 25 and JDK 27. Its README and `docs/output/` replace a `docs/` folder; the write-up is [the post](https://ankurm.com/jvm-flags-for-spring-boot-in-containers-jdk-25-27/) | Spring Boot 4.1.1, Docker, two JDKs |
|
| [`flags/`](flags) | The flag-by-flag production cheat sheet: `MaxRAMPercentage` vs a hardcoded `-Xmx` (one fails safely, one gets the container SIGKILLed), `-XX:-UseContainerSupport`, the `ActiveProcessorCount` lie, and which diagnostic flags (`HeapDumpOnOutOfMemoryError`, `NativeMemoryTracking`) are worth leaving on in prod — on real Docker containers, JDK 25 and JDK 27. Its README and `docs/output/` replace a `docs/` folder; the write-up is [the post](https://ankurm.com/jvm-flags-for-spring-boot-in-containers-jdk-25-27/) | Spring Boot 4.1.1, Docker, two JDKs |
|
||||||
| [`jdk28-zgc/`](jdk28-zgc) | ZGC startup and warmup versus initial heap size on JDK 25, 27 and 28 early access with a Spring Boot 4.1.1 server, plus a probe showing that JEP 545/546's `-XX:+ZAdaptiveHeapSizing` is not in any build yet. Its README and `output/` replace a `docs/` folder; the write-up is [the post](https://ankurm.com/zgc-jdk-28-adaptive-heap-sizing-faster-startup/) | Spring Boot 4.1.1, three JDKs |
|
| [`jdk28-zgc/`](jdk28-zgc) | ZGC startup and warmup versus initial heap size on JDK 25, 27 and 28 early access with a Spring Boot 4.1.1 server, plus a probe showing that JEP 545/546's `-XX:+ZAdaptiveHeapSizing` is not in any build yet. Its README and `output/` replace a `docs/` folder; the write-up is [the post](https://ankurm.com/zgc-jdk-28-adaptive-heap-sizing-faster-startup/) | Spring Boot 4.1.1, three JDKs |
|
||||||
|
|
||||||
|
|||||||
@@ -0,0 +1,62 @@
|
|||||||
|
# shenandoah-jdk28 — generational Shenandoah vs SATB Shenandoah vs G1 vs ZGC on a Spring Boot service
|
||||||
|
|
||||||
|
Companion module for the ankurm.com post
|
||||||
|
**Generational Shenandoah by Default in JDK 28 (JEP 535): Benchmarked Against G1 and ZGC**.
|
||||||
|
There is no `docs/` folder: this README and the captured files in [`output/`](output) are the documentation.
|
||||||
|
|
||||||
|
## The honest status
|
||||||
|
|
||||||
|
JEP 535 ("Shenandoah: Generational Mode by Default", which also deprecates the SATB mode) is **Targeted** to JDK 28 on
|
||||||
|
openjdk.org, but it is **not integrated in Temurin 28+21 EA**. [`output/01-gc-mode-probe.txt`](output/01-gc-mode-probe.txt)
|
||||||
|
shows `-XX:+UseShenandoahGC` still starts in *Snapshot-At-The-Beginning (SATB)* mode on both JDK 25.0.5+8 and 28+21.
|
||||||
|
Generational mode is available on both as an explicit opt-in (`-XX:ShenandoahGCMode=generational`).
|
||||||
|
So this module benchmarks **explicit generational vs explicit SATB**, which is what the default will switch between.
|
||||||
|
Oracle's jdk.java.net builds do not contain Shenandoah at all; use Temurin (or another build that ships it).
|
||||||
|
|
||||||
|
## Versions
|
||||||
|
|
||||||
|
| Component | Version |
|
||||||
|
|---|---|
|
||||||
|
| JDK 25 | Temurin 25.0.5+8 (probe only) |
|
||||||
|
| JDK 28 | Temurin 28+21 EA (28-beta+21-ea, build 202610221655) |
|
||||||
|
| Spring Boot | 4.1.2 (`spring-boot-starter-webmvc`) |
|
||||||
|
| Machine | 2 vCPU, 8 GB RAM Linux sandbox. A small box: absolute numbers are illustrative |
|
||||||
|
| Heap | `-Xms1g -Xmx1g` for every configuration |
|
||||||
|
|
||||||
|
## Quick start
|
||||||
|
|
||||||
|
```console
|
||||||
|
mvn -B -DskipTests package
|
||||||
|
export J25=/path/to/temurin-25 J28=/path/to/temurin-28-ea
|
||||||
|
scripts/run-all.sh # probe + 3 interleaved reps of 4 configurations, ~2.5 min per run
|
||||||
|
scripts/more-reps.sh 4 5 # append more reps and re-summarise
|
||||||
|
```
|
||||||
|
|
||||||
|
## What is measured
|
||||||
|
|
||||||
|
| File | What it is |
|
||||||
|
|---|---|
|
||||||
|
| [`src/main/java/com/ankurm/shenandoah/ShenandoahApplication.java`](src/main/java/com/ankurm/shenandoah/ShenandoahApplication.java) | `/work` replaces 8 of 200,000 1 KiB cache slots and allocates 100 short-lived 1 KiB arrays; `/info` reports the heap |
|
||||||
|
| [`scripts/LoadGen.java`](scripts/LoadGen.java) | `open`: fixed-rate load (3,000 req/s, 64 virtual threads), latency measured from the *intended* send time; `closed`: peak throughput |
|
||||||
|
| [`scripts/bench-one.sh`](scripts/bench-one.sh) | One run: 30 s warm-up, 60 s open-loop window, 20 s closed loop; pauses from `-Xlog:gc*`; peak RSS from `VmHWM` |
|
||||||
|
| [`scripts/gc_pauses.py`](scripts/gc_pauses.py) | Stop-the-world pauses inside the measure window |
|
||||||
|
| [`scripts/summarize.py`](scripts/summarize.py) | Median per configuration |
|
||||||
|
|
||||||
|
## Results (median of 5 interleaved runs, JDK 28+21 EA)
|
||||||
|
|
||||||
|
[`output/03-summary.txt`](output/03-summary.txt), from the per-run lines in [`output/02-runs.txt`](output/02-runs.txt):
|
||||||
|
|
||||||
|
| Configuration | p50 ms | p99 ms | p99.9 ms | max ms | peak req/s | pause max ms | RSS MB |
|
||||||
|
|---|---|---|---|---|---|---|---|
|
||||||
|
| G1 | 0.63 | 29.2 | 128.1 | 179.5 | 8,900 | 111.6 | 1,205 |
|
||||||
|
| ZGC (generational) | 0.60 | 7.4 | 17.3 | 36.4 | 7,117 | 0.17 | 1,250 |
|
||||||
|
| Shenandoah SATB | 0.78 | 14.9 | 43.6 | 86.4 | 5,026 | 1.5 | 1,212 |
|
||||||
|
| Shenandoah generational | 0.82 | 1,557.5 | 1,820.4 | 1,864.7 | 5,498 | 2.6 | 1,227 |
|
||||||
|
|
||||||
|
**Read the last row carefully.** Per-run p99 for generational Shenandoah was 1,557, 3,077, 24.5, 1,658 and 692 ms: four of five
|
||||||
|
runs contain a multi-second latency episode while the recorded stop-the-world pauses stay at a few milliseconds
|
||||||
|
(one run had a 209 ms `Pause Full`). The GC logs do not show allocation failures (`Trigger: Failed to allocate` count is 0), and
|
||||||
|
the `Cancelling GC: Concurrent GC` lines are old-generation marking being preempted for a young collection, which is normal.
|
||||||
|
So the cause of the stalls is **not established** here. One SATB run (811 ms p99) also stalled with no GC event at all,
|
||||||
|
which says this 2-vCPU machine produces stalls of its own. Treat the generational result as "did not behave on this box
|
||||||
|
and this workload", not as a general verdict. A larger machine and a bigger heap are the obvious next experiment.
|
||||||
@@ -0,0 +1,20 @@
|
|||||||
|
== JDK 25: OpenJDK 64-Bit Server VM Temurin-25.0.5+8 (build 25.0.5+8-LTS, mixed mode, sharing)
|
||||||
|
-- java -XX:+UseShenandoahGC -Xlog:gc+init -version
|
||||||
|
[0.009s][info][gc,init] Mode: Snapshot-At-The-Beginning (SATB)
|
||||||
|
-- java -XX:+UseShenandoahGC -XX:+PrintFlagsFinal -version (ShenandoahGCMode)
|
||||||
|
ccstr ShenandoahGCMode = satb {product} {default}
|
||||||
|
-- java -XX:+UseShenandoahGC -XX:ShenandoahGCMode=generational -Xlog:gc+init -version
|
||||||
|
[0.009s][info][gc,init] Mode: Generational
|
||||||
|
-- java -XX:+UseShenandoahGC -XX:ShenandoahGCMode=satb -Xlog:gc+init -version
|
||||||
|
[0.007s][info][gc,init] Mode: Snapshot-At-The-Beginning (SATB)
|
||||||
|
|
||||||
|
== JDK 28 EA: OpenJDK 64-Bit Server VM Temurin-28+21-202610221655 (build 28-beta+21-ea, mixed mode, sharing)
|
||||||
|
-- java -XX:+UseShenandoahGC -Xlog:gc+init -version
|
||||||
|
[0.006s][info][gc,init] Mode: Snapshot-At-The-Beginning (SATB)
|
||||||
|
-- java -XX:+UseShenandoahGC -XX:+PrintFlagsFinal -version (ShenandoahGCMode)
|
||||||
|
ccstr ShenandoahGCMode = satb {product} {default}
|
||||||
|
-- java -XX:+UseShenandoahGC -XX:ShenandoahGCMode=generational -Xlog:gc+init -version
|
||||||
|
[0.008s][info][gc,init] Mode: Generational
|
||||||
|
-- java -XX:+UseShenandoahGC -XX:ShenandoahGCMode=satb -Xlog:gc+init -version
|
||||||
|
[0.007s][info][gc,init] Mode: Snapshot-At-The-Beginning (SATB)
|
||||||
|
|
||||||
@@ -0,0 +1,20 @@
|
|||||||
|
label=g1 rep=1 target=3000 achieved=3000.0 requests=180000 p50_ms=0.61 p99_ms=28.69 p999_ms=95.48 max_ms=154.41 errors=0 peak_rps=8399.5 pauses=65 pause_p50_ms=23.459 pause_p99_ms=115.550 pause_max_ms=115.550 pause_total_ms=1349.5 rss_peak_mb=1205 heap_committed_mb=1024
|
||||||
|
label=zgc rep=1 target=3000 achieved=3000.0 requests=180000 p50_ms=0.60 p99_ms=7.32 p999_ms=65.10 max_ms=108.85 errors=0 peak_rps=7540.4 pauses=158 pause_p50_ms=0.014 pause_p99_ms=0.060 pause_max_ms=0.141 pause_total_ms=2.6 rss_peak_mb=1250 heap_committed_mb=1024
|
||||||
|
label=shenandoah-satb rep=1 target=3000 achieved=3000.0 requests=180000 p50_ms=0.80 p99_ms=14.94 p999_ms=29.12 max_ms=53.70 errors=0 peak_rps=5025.8 pauses=204 pause_p50_ms=0.165 pause_p99_ms=0.945 pause_max_ms=1.125 pause_total_ms=44.1 rss_peak_mb=1213 heap_committed_mb=1024
|
||||||
|
label=shenandoah-generational rep=1 target=3000 achieved=3000.0 requests=180000 p50_ms=0.89 p99_ms=1557.51 p999_ms=1866.05 max_ms=1906.45 errors=0 peak_rps=5484.3 pauses=319 pause_p50_ms=0.170 pause_p99_ms=1.236 pause_max_ms=1.418 pause_total_ms=76.5 rss_peak_mb=1227 heap_committed_mb=1024
|
||||||
|
label=g1 rep=2 target=3000 achieved=3000.0 requests=180000 p50_ms=0.66 p99_ms=29.17 p999_ms=128.06 max_ms=179.49 errors=0 peak_rps=8899.7 pauses=63 pause_p50_ms=24.557 pause_p99_ms=110.232 pause_max_ms=110.232 pause_total_ms=1366.7 rss_peak_mb=1207 heap_committed_mb=1024
|
||||||
|
label=zgc rep=2 target=3000 achieved=3000.0 requests=180000 p50_ms=0.60 p99_ms=6.98 p999_ms=15.80 max_ms=33.36 errors=0 peak_rps=7060.8 pauses=154 pause_p50_ms=0.014 pause_p99_ms=0.090 pause_max_ms=0.183 pause_total_ms=2.6 rss_peak_mb=1241 heap_committed_mb=1024
|
||||||
|
label=shenandoah-satb rep=2 target=3000 achieved=3000.0 requests=180000 p50_ms=0.78 p99_ms=14.58 p999_ms=27.28 max_ms=56.25 errors=0 peak_rps=4542.2 pauses=204 pause_p50_ms=0.160 pause_p99_ms=0.981 pause_max_ms=1.067 pause_total_ms=46.0 rss_peak_mb=1212 heap_committed_mb=1024
|
||||||
|
label=shenandoah-generational rep=2 target=3000 achieved=3000.0 requests=180000 p50_ms=0.93 p99_ms=3077.07 p999_ms=3222.87 max_ms=3284.93 errors=0 peak_rps=5497.7 pauses=328 pause_p50_ms=0.184 pause_p99_ms=1.544 pause_max_ms=209.368 pause_total_ms=292.7 rss_peak_mb=1219 heap_committed_mb=1024
|
||||||
|
label=g1 rep=3 target=3000 achieved=3000.0 requests=180000 p50_ms=0.65 p99_ms=29.18 p999_ms=137.90 max_ms=194.77 errors=0 peak_rps=9385.5 pauses=65 pause_p50_ms=24.353 pause_p99_ms=123.149 pause_max_ms=123.149 pause_total_ms=1383.9 rss_peak_mb=1203 heap_committed_mb=1024
|
||||||
|
label=zgc rep=3 target=3000 achieved=3000.0 requests=180000 p50_ms=0.62 p99_ms=7.37 p999_ms=15.73 max_ms=36.38 errors=0 peak_rps=7816.9 pauses=156 pause_p50_ms=0.014 pause_p99_ms=0.099 pause_max_ms=0.184 pause_total_ms=2.6 rss_peak_mb=1257 heap_committed_mb=1024
|
||||||
|
label=shenandoah-satb rep=3 target=3000 achieved=3000.0 requests=180000 p50_ms=0.69 p99_ms=811.54 p999_ms=1116.58 max_ms=1186.35 errors=0 peak_rps=5402.7 pauses=213 pause_p50_ms=0.144 pause_p99_ms=1.012 pause_max_ms=2.851 pause_total_ms=43.8 rss_peak_mb=1212 heap_committed_mb=1024
|
||||||
|
label=shenandoah-generational rep=3 target=3000 achieved=3000.0 requests=180000 p50_ms=0.82 p99_ms=24.49 p999_ms=103.01 max_ms=150.61 errors=0 peak_rps=5270.0 pauses=282 pause_p50_ms=0.172 pause_p99_ms=1.398 pause_max_ms=2.988 pause_total_ms=72.0 rss_peak_mb=1229 heap_committed_mb=1024
|
||||||
|
label=g1 rep=4 target=3000 achieved=3000.0 requests=180000 p50_ms=0.63 p99_ms=42.72 p999_ms=117.79 max_ms=146.07 errors=0 peak_rps=8847.7 pauses=55 pause_p50_ms=25.266 pause_p99_ms=111.576 pause_max_ms=111.576 pause_total_ms=1693.4 rss_peak_mb=1205 heap_committed_mb=1024
|
||||||
|
label=zgc rep=4 target=3000 achieved=3000.0 requests=180000 p50_ms=0.67 p99_ms=7.98 p999_ms=17.28 max_ms=31.52 errors=0 peak_rps=6923.5 pauses=155 pause_p50_ms=0.015 pause_p99_ms=0.097 pause_max_ms=0.168 pause_total_ms=3.0 rss_peak_mb=1250 heap_committed_mb=1024
|
||||||
|
label=shenandoah-satb rep=4 target=3000 achieved=3000.0 requests=180000 p50_ms=0.87 p99_ms=21.25 p999_ms=43.63 max_ms=86.39 errors=0 peak_rps=4639.2 pauses=210 pause_p50_ms=0.163 pause_p99_ms=0.743 pause_max_ms=1.475 pause_total_ms=44.1 rss_peak_mb=1209 heap_committed_mb=1024
|
||||||
|
label=shenandoah-generational rep=4 target=3000 achieved=3000.0 requests=180000 p50_ms=0.75 p99_ms=1657.87 p999_ms=1820.38 max_ms=1864.68 errors=0 peak_rps=5519.6 pauses=293 pause_p50_ms=0.177 pause_p99_ms=2.230 pause_max_ms=2.553 pause_total_ms=76.6 rss_peak_mb=1220 heap_committed_mb=1024
|
||||||
|
label=g1 rep=5 target=3000 achieved=3000.0 requests=180000 p50_ms=0.60 p99_ms=66.31 p999_ms=379.56 max_ms=423.34 errors=0 peak_rps=9775.1 pauses=70 pause_p50_ms=25.645 pause_p99_ms=101.489 pause_max_ms=101.489 pause_total_ms=1443.9 rss_peak_mb=1202 heap_committed_mb=1024
|
||||||
|
label=zgc rep=5 target=3000 achieved=3000.0 requests=180000 p50_ms=0.59 p99_ms=7.49 p999_ms=48.76 max_ms=83.78 errors=0 peak_rps=7117.4 pauses=147 pause_p50_ms=0.014 pause_p99_ms=0.080 pause_max_ms=0.112 pause_total_ms=2.4 rss_peak_mb=1243 heap_committed_mb=1024
|
||||||
|
label=shenandoah-satb rep=5 target=3000 achieved=3000.0 requests=180000 p50_ms=0.64 p99_ms=12.71 p999_ms=62.14 max_ms=101.88 errors=0 peak_rps=5448.0 pauses=191 pause_p50_ms=0.147 pause_p99_ms=1.192 pause_max_ms=1.512 pause_total_ms=36.9 rss_peak_mb=1214 heap_committed_mb=1024
|
||||||
|
label=shenandoah-generational rep=5 target=3000 achieved=3000.0 requests=180000 p50_ms=0.72 p99_ms=691.58 p999_ms=916.21 max_ms=959.93 errors=0 peak_rps=5745.0 pauses=277 pause_p50_ms=0.176 pause_p99_ms=1.283 pause_max_ms=1.359 pause_total_ms=67.2 rss_peak_mb=1231 heap_committed_mb=1024
|
||||||
@@ -0,0 +1,7 @@
|
|||||||
|
median of 5 runs per configuration (latency from a fixed-rate open loop; throughput from a closed loop)
|
||||||
|
|
||||||
|
configuration p50 ms p99 ms p99.9 ms max ms peak req/s pauses pause p99 ms pause max ms pause total ms RSS MB
|
||||||
|
g1 0.63 29.2 128.1 179.5 8900 65 111.576 111.576 1384 1205
|
||||||
|
zgc 0.60 7.4 17.3 36.4 7117 155 0.090 0.168 3 1250
|
||||||
|
shenandoah-satb 0.78 14.9 43.6 86.4 5026 204 0.981 1.475 44 1212
|
||||||
|
shenandoah-generational 0.82 1557.5 1820.4 1864.7 5498 293 1.398 2.553 76 1227
|
||||||
@@ -0,0 +1,33 @@
|
|||||||
|
<?xml version="1.0" encoding="UTF-8"?>
|
||||||
|
<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
|
||||||
|
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 https://maven.apache.org/xsd/maven-4.0.0.xsd">
|
||||||
|
<modelVersion>4.0.0</modelVersion>
|
||||||
|
<parent>
|
||||||
|
<groupId>org.springframework.boot</groupId>
|
||||||
|
<artifactId>spring-boot-starter-parent</artifactId>
|
||||||
|
<version>4.1.2</version>
|
||||||
|
<relativePath/>
|
||||||
|
</parent>
|
||||||
|
<groupId>com.ankurm</groupId>
|
||||||
|
<artifactId>shenandoah-jdk28</artifactId>
|
||||||
|
<version>1.0.0</version>
|
||||||
|
<name>shenandoah-jdk28</name>
|
||||||
|
<description>Same Spring Boot load on G1, generational ZGC and Shenandoah (SATB and generational) on JDK 28 early access</description>
|
||||||
|
<properties>
|
||||||
|
<java.version>25</java.version>
|
||||||
|
</properties>
|
||||||
|
<dependencies>
|
||||||
|
<dependency>
|
||||||
|
<groupId>org.springframework.boot</groupId>
|
||||||
|
<artifactId>spring-boot-starter-webmvc</artifactId>
|
||||||
|
</dependency>
|
||||||
|
</dependencies>
|
||||||
|
<build>
|
||||||
|
<plugins>
|
||||||
|
<plugin>
|
||||||
|
<groupId>org.springframework.boot</groupId>
|
||||||
|
<artifactId>spring-boot-maven-plugin</artifactId>
|
||||||
|
</plugin>
|
||||||
|
</plugins>
|
||||||
|
</build>
|
||||||
|
</project>
|
||||||
@@ -0,0 +1,94 @@
|
|||||||
|
import java.net.URI;
|
||||||
|
import java.net.http.HttpClient;
|
||||||
|
import java.net.http.HttpRequest;
|
||||||
|
import java.net.http.HttpResponse;
|
||||||
|
import java.time.Duration;
|
||||||
|
import java.util.Arrays;
|
||||||
|
import java.util.concurrent.ExecutorService;
|
||||||
|
import java.util.concurrent.Executors;
|
||||||
|
import java.util.concurrent.TimeUnit;
|
||||||
|
import java.util.concurrent.atomic.AtomicInteger;
|
||||||
|
import java.util.concurrent.atomic.AtomicLong;
|
||||||
|
import java.util.concurrent.locks.LockSupport;
|
||||||
|
|
||||||
|
/**
|
||||||
|
* Load generator, run with the source launcher: java LoadGen.java open|closed URL RATE_OR_0 CONCURRENCY WARMUP_S MEASURE_S
|
||||||
|
*
|
||||||
|
* open : requests are SCHEDULED at a fixed rate and latency is measured from the scheduled time, so a server
|
||||||
|
* that stalls is charged for the requests that queued behind the stall (no coordinated omission).
|
||||||
|
* closed : CONCURRENCY clients send back to back; the result is the peak throughput.
|
||||||
|
*
|
||||||
|
* Prints one line: mode=... achieved=... p50_ms=... p99_ms=... p999_ms=... max_ms=... errors=... window_start_ms=... window_end_ms=...
|
||||||
|
* (the window is the measured interval in epoch milliseconds, so GC pauses can be cut to exactly that interval).
|
||||||
|
*/
|
||||||
|
public class LoadGen {
|
||||||
|
|
||||||
|
public static void main(String[] a) throws Exception {
|
||||||
|
String mode = a[0];
|
||||||
|
URI uri = URI.create(a[1]);
|
||||||
|
double rate = Double.parseDouble(a[2]);
|
||||||
|
int concurrency = Integer.parseInt(a[3]);
|
||||||
|
int warmup = Integer.parseInt(a[4]);
|
||||||
|
int measure = Integer.parseInt(a[5]);
|
||||||
|
boolean open = mode.equals("open");
|
||||||
|
|
||||||
|
HttpClient client = HttpClient.newBuilder().version(HttpClient.Version.HTTP_1_1)
|
||||||
|
.connectTimeout(Duration.ofSeconds(5)).build();
|
||||||
|
HttpRequest request = HttpRequest.newBuilder(uri).timeout(Duration.ofSeconds(30)).GET().build();
|
||||||
|
|
||||||
|
long start = System.nanoTime();
|
||||||
|
long epochStartMs = System.currentTimeMillis();
|
||||||
|
long measureFrom = start + warmup * 1_000_000_000L;
|
||||||
|
long end = measureFrom + measure * 1_000_000_000L;
|
||||||
|
int capacity = (int) (open ? rate * measure * 2 + 1000 : 2_000_000);
|
||||||
|
long[] latencies = new long[capacity];
|
||||||
|
AtomicInteger count = new AtomicInteger();
|
||||||
|
AtomicLong errors = new AtomicLong();
|
||||||
|
AtomicLong next = new AtomicLong();
|
||||||
|
long intervalNanos = open ? (long) (1e9 / rate) : 0;
|
||||||
|
|
||||||
|
ExecutorService pool = Executors.newVirtualThreadPerTaskExecutor();
|
||||||
|
for (int t = 0; t < concurrency; t++) {
|
||||||
|
pool.submit(() -> {
|
||||||
|
while (true) {
|
||||||
|
long intended;
|
||||||
|
if (open) {
|
||||||
|
intended = start + next.getAndIncrement() * intervalNanos;
|
||||||
|
if (intended >= end) return;
|
||||||
|
long wait = intended - System.nanoTime();
|
||||||
|
if (wait > 0) LockSupport.parkNanos(wait);
|
||||||
|
} else {
|
||||||
|
intended = System.nanoTime();
|
||||||
|
if (intended >= end) return;
|
||||||
|
}
|
||||||
|
try {
|
||||||
|
HttpResponse<Void> r = client.send(request, HttpResponse.BodyHandlers.discarding());
|
||||||
|
long done = System.nanoTime();
|
||||||
|
if (r.statusCode() != 200) errors.incrementAndGet();
|
||||||
|
if (intended >= measureFrom) {
|
||||||
|
int i = count.getAndIncrement();
|
||||||
|
if (i < latencies.length) latencies[i] = done - intended;
|
||||||
|
}
|
||||||
|
} catch (Exception e) {
|
||||||
|
errors.incrementAndGet();
|
||||||
|
}
|
||||||
|
}
|
||||||
|
});
|
||||||
|
}
|
||||||
|
pool.shutdown();
|
||||||
|
pool.awaitTermination(measure + warmup + 120, TimeUnit.SECONDS);
|
||||||
|
|
||||||
|
int n = Math.min(count.get(), latencies.length);
|
||||||
|
long[] sorted = Arrays.copyOf(latencies, n);
|
||||||
|
Arrays.sort(sorted);
|
||||||
|
System.out.printf("mode=%s target=%.0f achieved=%.1f requests=%d p50_ms=%.2f p99_ms=%.2f p999_ms=%.2f max_ms=%.2f errors=%d window_start_ms=%d window_end_ms=%d%n",
|
||||||
|
mode, rate, n / (double) measure, n, ms(sorted, 0.50), ms(sorted, 0.99), ms(sorted, 0.999),
|
||||||
|
n == 0 ? 0 : sorted[n - 1] / 1e6, errors.get(),
|
||||||
|
epochStartMs + warmup * 1000L, epochStartMs + (warmup + measure) * 1000L);
|
||||||
|
}
|
||||||
|
|
||||||
|
static double ms(long[] sorted, double q) {
|
||||||
|
if (sorted.length == 0) return 0;
|
||||||
|
return sorted[Math.min(sorted.length - 1, (int) Math.ceil(q * sorted.length) - 1)] / 1e6;
|
||||||
|
}
|
||||||
|
}
|
||||||
Executable
+35
@@ -0,0 +1,35 @@
|
|||||||
|
#!/usr/bin/env bash
|
||||||
|
# One benchmark run: start the server with the given JVM flags, warm up, measure at a fixed request rate
|
||||||
|
# (latency), then measure peak throughput, and print ONE result line.
|
||||||
|
#
|
||||||
|
# bench-one.sh <java-home> <label> <rep> [jvm flags...]
|
||||||
|
#
|
||||||
|
# Env: RATE (default 3000 req/s), WARMUP (30 s), MEASURE (60 s), PORT (18090)
|
||||||
|
set -u
|
||||||
|
JH=$1; LABEL=$2; REP=$3; shift 3
|
||||||
|
HERE="$(cd "$(dirname "$0")" && pwd)"
|
||||||
|
JAR="$HERE/../target/shenandoah-jdk28-1.0.0.jar"
|
||||||
|
RATE=${RATE:-3000}; WARMUP=${WARMUP:-30}; MEASURE=${MEASURE:-60}; PORT=${PORT:-18090}
|
||||||
|
OUT="$HERE/../target/gclogs"; mkdir -p "$OUT"
|
||||||
|
LOG="$OUT/$LABEL-$REP.log"
|
||||||
|
|
||||||
|
env -u JAVA_TOOL_OPTIONS "$JH/bin/java" -Xms1g -Xmx1g "$@" \
|
||||||
|
-Xlog:gc*:file="$LOG":time,uptime \
|
||||||
|
-jar "$JAR" --server.port=$PORT --spring.main.banner-mode=off --logging.level.root=WARN >/dev/null 2>&1 &
|
||||||
|
PID=$!
|
||||||
|
for _ in $(seq 1 200); do
|
||||||
|
curl -s -o /dev/null -m 1 "localhost:$PORT/work" && break
|
||||||
|
kill -0 $PID 2>/dev/null || { echo "label=$LABEL rep=$REP FAILED: JVM exited"; exit 1; }
|
||||||
|
sleep 0.3
|
||||||
|
done
|
||||||
|
|
||||||
|
OPEN=$(env -u JAVA_TOOL_OPTIONS "$JH/bin/java" "$HERE/LoadGen.java" open "http://localhost:$PORT/work" "$RATE" 64 "$WARMUP" "$MEASURE")
|
||||||
|
CLOSED=$(env -u JAVA_TOOL_OPTIONS "$JH/bin/java" "$HERE/LoadGen.java" closed "http://localhost:$PORT/work" 0 16 10 20)
|
||||||
|
RSS_PEAK_MB=$(awk '/VmHWM/ {printf "%d", $2/1024}' /proc/$PID/status)
|
||||||
|
COMMITTED=$(curl -s "localhost:$PORT/info" | sed -n 's/.*"committedMiB":\([0-9]*\).*/\1/p')
|
||||||
|
kill $PID 2>/dev/null; wait $PID 2>/dev/null
|
||||||
|
|
||||||
|
W0=$(sed -n 's/.*window_start_ms=\([0-9]*\).*/\1/p' <<<"$OPEN"); W1=$(sed -n 's/.*window_end_ms=\([0-9]*\).*/\1/p' <<<"$OPEN")
|
||||||
|
PAUSES=$(python3 "$HERE/gc_pauses.py" "$LOG" "$W0" "$W1")
|
||||||
|
CLOSED_TPUT=$(sed -n 's/.*achieved=\([0-9.]*\).*/\1/p' <<<"$CLOSED")
|
||||||
|
echo "label=$LABEL rep=$REP $(sed 's/ window_start_ms=.*//; s/^mode=open //' <<<"$OPEN") peak_rps=$CLOSED_TPUT $PAUSES rss_peak_mb=$RSS_PEAK_MB heap_committed_mb=$COMMITTED"
|
||||||
Executable
+33
@@ -0,0 +1,33 @@
|
|||||||
|
#!/usr/bin/env python3
|
||||||
|
"""Stop-the-world pauses from a -Xlog:gc*:...:time,uptime log, restricted to an epoch-ms window.
|
||||||
|
|
||||||
|
usage: gc_pauses.py LOG WINDOW_START_MS WINDOW_END_MS
|
||||||
|
|
||||||
|
A pause is a log line that contains "Pause" and ends in "<n>ms" (G1 "Pause Young ... 4.2ms", ZGC
|
||||||
|
"Y: Pause Mark Start 0.018ms", Shenandoah "Pause Init Mark 0.052ms"). The line without a duration that
|
||||||
|
G1 and Shenandoah print when a pause *starts* is ignored, so nothing is counted twice.
|
||||||
|
"""
|
||||||
|
import re
|
||||||
|
import sys
|
||||||
|
from datetime import datetime
|
||||||
|
|
||||||
|
LINE = re.compile(r"^\[(\d{4}-\d\d-\d\dT[\d:.]+[+-]\d{4})\].*?\bGC\(\d+\) .*Pause .* ([0-9.]+)ms$")
|
||||||
|
|
||||||
|
log, lo, hi = sys.argv[1], int(sys.argv[2]), int(sys.argv[3])
|
||||||
|
pauses = []
|
||||||
|
for line in open(log, errors="replace"):
|
||||||
|
m = LINE.match(line.rstrip("\n"))
|
||||||
|
if not m:
|
||||||
|
continue
|
||||||
|
ts = datetime.strptime(m.group(1), "%Y-%m-%dT%H:%M:%S.%f%z").timestamp() * 1000
|
||||||
|
if lo <= ts <= hi:
|
||||||
|
pauses.append(float(m.group(2)))
|
||||||
|
pauses.sort()
|
||||||
|
n = len(pauses)
|
||||||
|
|
||||||
|
|
||||||
|
def q(p):
|
||||||
|
return pauses[min(n - 1, int(p * n))] if n else 0.0
|
||||||
|
|
||||||
|
|
||||||
|
print(f"pauses={n} pause_p50_ms={q(0.50):.3f} pause_p99_ms={q(0.99):.3f} pause_max_ms={pauses[-1] if n else 0.0:.3f} pause_total_ms={sum(pauses):.1f}")
|
||||||
Executable
+13
@@ -0,0 +1,13 @@
|
|||||||
|
#!/usr/bin/env bash
|
||||||
|
# Appends extra interleaved repetitions to output/02-runs.txt (same protocol as run-all.sh), then re-summarises.
|
||||||
|
# scripts/more-reps.sh FIRST_REP LAST_REP e.g. scripts/more-reps.sh 4 5
|
||||||
|
set -eu
|
||||||
|
cd "$(dirname "$0")/.."
|
||||||
|
: "${J28:?set J28}"
|
||||||
|
for rep in $(seq "$1" "$2"); do
|
||||||
|
./scripts/bench-one.sh "$J28" g1 "$rep" -XX:+UseG1GC | tee -a output/02-runs.txt
|
||||||
|
./scripts/bench-one.sh "$J28" zgc "$rep" -XX:+UseZGC | tee -a output/02-runs.txt
|
||||||
|
./scripts/bench-one.sh "$J28" shenandoah-satb "$rep" -XX:+UseShenandoahGC -XX:ShenandoahGCMode=satb | tee -a output/02-runs.txt
|
||||||
|
./scripts/bench-one.sh "$J28" shenandoah-generational "$rep" -XX:+UseShenandoahGC -XX:ShenandoahGCMode=generational | tee -a output/02-runs.txt
|
||||||
|
done
|
||||||
|
python3 scripts/summarize.py output/02-runs.txt | tee output/03-summary.txt
|
||||||
Executable
+17
@@ -0,0 +1,17 @@
|
|||||||
|
#!/usr/bin/env bash
|
||||||
|
# What does plain -XX:+UseShenandoahGC select on each JDK? Needs J25 and J28 pointing at JDKs that include Shenandoah
|
||||||
|
# (Temurin does; the Oracle builds from jdk.java.net do not).
|
||||||
|
set -u
|
||||||
|
for entry in "JDK 25:$J25" "JDK 28 EA:$J28"; do
|
||||||
|
name=${entry%%:*}; home=${entry#*:}
|
||||||
|
echo "== $name: $(env -u JAVA_TOOL_OPTIONS "$home/bin/java" -version 2>&1 | sed -n 3p)"
|
||||||
|
echo "-- java -XX:+UseShenandoahGC -Xlog:gc+init -version"
|
||||||
|
env -u JAVA_TOOL_OPTIONS "$home/bin/java" -XX:+UseShenandoahGC -Xlog:gc+init -Xmx256m -version 2>&1 | grep -E 'Mode:|warning|WARNING|deprecat'
|
||||||
|
echo "-- java -XX:+UseShenandoahGC -XX:+PrintFlagsFinal -version (ShenandoahGCMode)"
|
||||||
|
env -u JAVA_TOOL_OPTIONS "$home/bin/java" -XX:+UseShenandoahGC -XX:+PrintFlagsFinal -Xmx256m -version 2>&1 | grep -E 'ShenandoahGCMode'
|
||||||
|
echo "-- java -XX:+UseShenandoahGC -XX:ShenandoahGCMode=generational -Xlog:gc+init -version"
|
||||||
|
env -u JAVA_TOOL_OPTIONS "$home/bin/java" -XX:+UseShenandoahGC -XX:ShenandoahGCMode=generational -Xlog:gc+init -Xmx256m -version 2>&1 | grep -E 'Mode:|warning|WARNING|deprecat'
|
||||||
|
echo "-- java -XX:+UseShenandoahGC -XX:ShenandoahGCMode=satb -Xlog:gc+init -version"
|
||||||
|
env -u JAVA_TOOL_OPTIONS "$home/bin/java" -XX:+UseShenandoahGC -XX:ShenandoahGCMode=satb -Xlog:gc+init -Xmx256m -version 2>&1 | grep -E 'Mode:|warning|WARNING|deprecat'
|
||||||
|
echo
|
||||||
|
done
|
||||||
Executable
+21
@@ -0,0 +1,21 @@
|
|||||||
|
#!/usr/bin/env bash
|
||||||
|
# Regenerates everything in output/.
|
||||||
|
# export J25=/path/to/temurin-25 J28=/path/to/temurin-28-ea (both must include Shenandoah: use Temurin, not the jdk.java.net builds)
|
||||||
|
# mvn -B -DskipTests package
|
||||||
|
# scripts/run-all.sh # 4 configurations x REPS runs, interleaved, about 2.5 minutes per run
|
||||||
|
#
|
||||||
|
# Env: REPS (default 3), RATE / WARMUP / MEASURE (see bench-one.sh)
|
||||||
|
set -eu
|
||||||
|
cd "$(dirname "$0")/.."
|
||||||
|
: "${J25:?set J25}" "${J28:?set J28}"
|
||||||
|
REPS=${REPS:-3}
|
||||||
|
./scripts/probe.sh > output/01-gc-mode-probe.txt 2>&1
|
||||||
|
: > output/02-runs.txt
|
||||||
|
for rep in $(seq 1 "$REPS"); do
|
||||||
|
# interleaved so slow drift on the machine hits every configuration alike
|
||||||
|
./scripts/bench-one.sh "$J28" g1 "$rep" -XX:+UseG1GC | tee -a output/02-runs.txt
|
||||||
|
./scripts/bench-one.sh "$J28" zgc "$rep" -XX:+UseZGC | tee -a output/02-runs.txt
|
||||||
|
./scripts/bench-one.sh "$J28" shenandoah-satb "$rep" -XX:+UseShenandoahGC -XX:ShenandoahGCMode=satb | tee -a output/02-runs.txt
|
||||||
|
./scripts/bench-one.sh "$J28" shenandoah-generational "$rep" -XX:+UseShenandoahGC -XX:ShenandoahGCMode=generational | tee -a output/02-runs.txt
|
||||||
|
done
|
||||||
|
python3 scripts/summarize.py output/02-runs.txt | tee output/03-summary.txt
|
||||||
Executable
+33
@@ -0,0 +1,33 @@
|
|||||||
|
#!/usr/bin/env python3
|
||||||
|
"""Median across repetitions, one row per configuration. usage: summarize.py output/02-runs.txt"""
|
||||||
|
import re
|
||||||
|
import statistics
|
||||||
|
import sys
|
||||||
|
from collections import OrderedDict
|
||||||
|
|
||||||
|
rows = OrderedDict()
|
||||||
|
for line in open(sys.argv[1]):
|
||||||
|
if "FAILED" in line or not line.strip():
|
||||||
|
continue
|
||||||
|
kv = dict(re.findall(r"(\w+)=([^\s]+)", line))
|
||||||
|
rows.setdefault(kv["label"], []).append({k: float(v) for k, v in kv.items() if k not in ("label",)})
|
||||||
|
|
||||||
|
cols = [
|
||||||
|
("p50 ms", "p50_ms", "{:.2f}"),
|
||||||
|
("p99 ms", "p99_ms", "{:.1f}"),
|
||||||
|
("p99.9 ms", "p999_ms", "{:.1f}"),
|
||||||
|
("max ms", "max_ms", "{:.1f}"),
|
||||||
|
("peak req/s", "peak_rps", "{:.0f}"),
|
||||||
|
("pauses", "pauses", "{:.0f}"),
|
||||||
|
("pause p99 ms", "pause_p99_ms", "{:.3f}"),
|
||||||
|
("pause max ms", "pause_max_ms", "{:.3f}"),
|
||||||
|
("pause total ms", "pause_total_ms", "{:.0f}"),
|
||||||
|
("RSS MB", "rss_peak_mb", "{:.0f}"),
|
||||||
|
]
|
||||||
|
reps = max(len(v) for v in rows.values())
|
||||||
|
print(f"median of {reps} runs per configuration (latency from a fixed-rate open loop; throughput from a closed loop)")
|
||||||
|
print()
|
||||||
|
print(f"{'configuration':<26}" + "".join(f"{c[0]:>15}" for c in cols))
|
||||||
|
for label, runs in rows.items():
|
||||||
|
cells = [c[2].format(statistics.median(r[c[1]] for r in runs)) for c in cols]
|
||||||
|
print(f"{label:<26}" + "".join(f"{x:>15}" for x in cells))
|
||||||
@@ -0,0 +1,71 @@
|
|||||||
|
package com.ankurm.shenandoah;
|
||||||
|
|
||||||
|
import java.lang.management.ManagementFactory;
|
||||||
|
import java.util.LinkedHashMap;
|
||||||
|
import java.util.Map;
|
||||||
|
import java.util.concurrent.ThreadLocalRandom;
|
||||||
|
import java.util.concurrent.atomic.AtomicReferenceArray;
|
||||||
|
import org.springframework.boot.SpringApplication;
|
||||||
|
import org.springframework.boot.autoconfigure.SpringBootApplication;
|
||||||
|
import org.springframework.web.bind.annotation.GetMapping;
|
||||||
|
import org.springframework.web.bind.annotation.RestController;
|
||||||
|
|
||||||
|
/**
|
||||||
|
* A server with the shape a generational collector is built for: a large, slowly changing live set
|
||||||
|
* (the cache) plus a steady stream of short-lived garbage (each request).
|
||||||
|
*/
|
||||||
|
@SpringBootApplication
|
||||||
|
public class ShenandoahApplication {
|
||||||
|
|
||||||
|
public static void main(String[] args) {
|
||||||
|
SpringApplication.run(ShenandoahApplication.class, args);
|
||||||
|
}
|
||||||
|
|
||||||
|
@RestController
|
||||||
|
static class Api {
|
||||||
|
|
||||||
|
/** 200,000 slots of 1 KiB is a live set of roughly 210 MiB. */
|
||||||
|
static final int SLOTS = 200_000;
|
||||||
|
/** Slots replaced per request: the old value becomes old-generation garbage. */
|
||||||
|
static final int REPLACED_PER_REQUEST = 8;
|
||||||
|
/** Short-lived 1 KiB arrays allocated per request: about 100 KiB of young garbage. */
|
||||||
|
static final int TEMPORARIES = 100;
|
||||||
|
|
||||||
|
private final AtomicReferenceArray<byte[]> cache = new AtomicReferenceArray<>(SLOTS);
|
||||||
|
/** Written on every request so the JIT cannot discard the temporaries. */
|
||||||
|
private volatile int sink;
|
||||||
|
|
||||||
|
Api() {
|
||||||
|
for (int i = 0; i < SLOTS; i++) {
|
||||||
|
cache.set(i, new byte[1024]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
@GetMapping("/work")
|
||||||
|
String work() {
|
||||||
|
ThreadLocalRandom r = ThreadLocalRandom.current();
|
||||||
|
for (int i = 0; i < REPLACED_PER_REQUEST; i++) {
|
||||||
|
cache.set(r.nextInt(SLOTS), new byte[1024]);
|
||||||
|
}
|
||||||
|
byte[][] temp = new byte[TEMPORARIES][];
|
||||||
|
for (int i = 0; i < TEMPORARIES; i++) {
|
||||||
|
temp[i] = new byte[1024];
|
||||||
|
temp[i][0] = (byte) i;
|
||||||
|
}
|
||||||
|
sink = temp[r.nextInt(TEMPORARIES)][0] + cache.get(r.nextInt(SLOTS))[0];
|
||||||
|
return "ok";
|
||||||
|
}
|
||||||
|
|
||||||
|
@GetMapping("/info")
|
||||||
|
Map<String, Object> info() {
|
||||||
|
Runtime rt = Runtime.getRuntime();
|
||||||
|
Map<String, Object> m = new LinkedHashMap<>();
|
||||||
|
m.put("javaVersion", System.getProperty("java.version"));
|
||||||
|
m.put("committedMiB", rt.totalMemory() >> 20);
|
||||||
|
m.put("maxMiB", rt.maxMemory() >> 20);
|
||||||
|
m.put("gc", ManagementFactory.getGarbageCollectorMXBeans().stream().map(b -> b.getName()).toList());
|
||||||
|
m.put("jvmUptimeMs", ManagementFactory.getRuntimeMXBean().getUptime());
|
||||||
|
return m;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
Reference in New Issue
Block a user