jvm-cds-opportunity-for-repeated-startup

What It Does

Detects Java repositories where short-lived JVM startup likely dominates CI runtime and suggests evaluating CDS/AppCDS or related startup amortization techniques.

Why It Matters

Each short-lived JVM invocation spends a significant fraction of wall time on class loading, bytecode verification, and JIT warmup. When CI pipelines repeatedly launch independent JVM processes (e.g. per-test-class forks, Gradle test executors, Maven Surefire forks), this overhead adds up.

Class Data Sharing (CDS/AppCDS) lets the JVM archive resolved class metadata so subsequent launches skip much of the class-loading phase. For repeated JVM startup patterns in CI, this can meaningfully reduce total workflow duration.

Triggers

Repository evidence (any of):

• pom.xml (Maven)
• build.gradle / build.gradle.kts (Gradle)
• gradlew / gradlew.bat (Gradle wrapper)
• src/main/java or src/test/java (Java source)

Workflow evidence (2+ occurrences across workflows):

• mvn test, mvn verify, or other test-related Maven lifecycle commands
• gradle test, ./gradlew test
• java -jar ...
• spring-boot:run in CI

Evidence is limited to non-release JVM steps: either repeated test commands or repeated direct java -jar/java -cp launches. Ordinary builds alone do not prove repeated short-lived test JVMs. Matrix boosts come from the JVM job, and diagnostics point at a relevant JVM command. CDS flags in JAVA_TOOL_OPTIONS, JDK_JAVA_OPTIONS, GRADLE_OPTS and MAVEN_OPTS are recognized.

Nested build roots referenced by workflow working directories or inline cd participate in JVM detection.

Non-Triggers

• Repositories already using -Xshare, -XX:SharedArchiveFile, -XX:ArchiveClassesAtExit, -XX:DumpLoadedClassList, or other CDS/AppCDS flags
• Native-image-first repos (GraalVM native-image, Quarkus native builds)
• Repositories with fewer than 2 JVM workflow commands
• Release-only image builds without repeated JVM test startup
• Single long-running JVM execution (e.g. a single java -jar application server)

Severity

warning — exploratory performance opportunity

Score Boosters

Higher score when any of these are detected:

• Maven Surefire/Failsafe forkCount configuration
• reuseForks=false (creates fresh JVM per test class)
• Gradle maxParallelForks > 1
• Spring Boot dependency
• Integration / e2e / smoke test steps
• Matrix builds (multiplies JVM startup count)
• Repeated Java workflow steps across jobs
• Surefire or Failsafe plugin presence

Advisory

CDS/AppCDS introduces archive generation and setup cost. The benefit depends on:

• How many times similar JVM startup patterns repeat
• Archive generation overhead relative to total CI time
• Whether test environment supports archive reuse (ephemeral CI runners may need to regenerate per run)

Always measure before and after. Compare total workflow duration, not isolated JVM startup latency.