book

Optimizing Java

by Benjamin J. Evans, James Gough, Chris Newland

April 2018

Intermediate to advanced

440 pages

11h 15m

English

O'Reilly Media, Inc.

Book available

Read now

Unlock full access

Conventions Used in This BookUsing Code ExamplesO’Reilly SafariHow to Contact UsAcknowledgments
Java Performance—The Wrong WayJava Performance OverviewPerformance as an Experimental ScienceA Taxonomy for PerformanceThroughputLatencyCapacityUtilizationEfficiencyScalabilityDegradationConnections Between the ObservablesReading Performance GraphsSummary
Interpreting and ClassloadingExecuting BytecodeIntroducing HotSpotIntroducing Just-in-Time CompilationJVM Memory ManagementThreading and the Java Memory ModelMeet the JVMsA Note on LicensesMonitoring and Tooling for the JVMVisualVMSummary
Introduction to Modern HardwareMemoryMemory CachesModern Processor FeaturesTranslation Lookaside BufferBranch Prediction and Speculative ExecutionHardware Memory ModelsOperating SystemsThe SchedulerA Question of TimeContext SwitchesA Simple System ModelBasic Detection StrategiesUtilizing the CPUGarbage CollectionI/OMechanical SympathyVirtualizationThe JVM and the Operating SystemSummary
Types of Performance TestLatency TestThroughput TestLoad TestStress TestEndurance TestCapacity Planning TestDegradation TestBest Practices PrimerTop-Down PerformanceCreating a Test EnvironmentIdentifying Performance RequirementsJava-Specific IssuesPerformance Testing as Part of the SDLCIntroducing Performance AntipatternsBoredomRésumé PaddingPeer PressureLack of UnderstandingMisunderstood/Nonexistent ProblemPerformance Antipatterns CatalogueDistracted by ShinyDistracted by SimplePerformance Tuning WizardTuning by FolkloreThe Blame DonkeyMissing the Bigger PictureUAT Is My DesktopProduction-Like Data Is HardCognitive Biases and Performance TestingReductionist ThinkingConfirmation BiasFog of War (Action Bias)Risk BiasEllsberg’s ParadoxSummary
Introduction to Measuring Java PerformanceIntroduction to JMHDon’t Microbenchmark If You Can Help It (A True Story)Heuristics for When to MicrobenchmarkThe JMH FrameworkExecuting BenchmarksStatistics for JVM PerformanceTypes of ErrorNon-Normal StatisticsInterpretation of StatisticsSummary
Introducing Mark and SweepGarbage Collection GlossaryIntroducing the HotSpot RuntimeRepresenting Objects at RuntimeGC Roots and ArenasAllocation and LifetimeWeak Generational HypothesisGarbage Collection in HotSpotThread-Local AllocationHemispheric CollectionThe Parallel CollectorsYoung Parallel CollectionsOld Parallel CollectionsLimitations of Parallel CollectorsThe Role of AllocationSummary
Tradeoffs and Pluggable CollectorsConcurrent GC TheoryJVM SafepointsTri-Color MarkingCMSHow CMS WorksBasic JVM Flags for CMSG1G1 Heap Layout and RegionsG1 Algorithm DesignG1 PhasesBasic JVM Flags for G1ShenandoahConcurrent CompactionObtaining ShenandoahC4 (Azul Zing)The Loaded Value BarrierBalanced (IBM J9)J9 Object HeadersLarge Arrays in BalancedNUMA and BalancedLegacy HotSpot CollectorsSerial and SerialOldIncremental CMS (iCMS)Deprecated and Removed GC CombinationsEpsilonSummary
Introduction to GC LoggingSwitching On GC LoggingGC Logs Versus JMXDrawbacks of JMXBenefits of GC Log DataLog Parsing ToolsCensumGCViewerDifferent Visualizations of the Same DataBasic GC TuningUnderstanding AllocationUnderstanding Pause TimeCollector Threads and GC RootsTuning Parallel GCTuning CMSConcurrent Mode Failures Due to FragmentationTuning G1jHiccupSummary

Overview of Bytecode InterpretationIntroduction to JVM BytecodeSimple InterpretersHotSpot-Specific DetailsAOT and JIT CompilationAOT CompilationJIT CompilationComparing AOT and JIT CompilationHotSpot JIT BasicsKlass Words, Vtables, and Pointer SwizzlingLogging JIT CompilationCompilers Within HotSpotTiered Compilation in HotSpotThe Code CacheFragmentationSimple JIT TuningSummary
Introducing JITWatchBasic JITWatch ViewsDebug JVMs and hsdisIntroducing JIT CompilationInliningInlining LimitsTuning the Inlining SubsystemLoop UnrollingLoop Unrolling SummaryEscape AnalysisEliminating Heap AllocationsLocks and Escape AnalysisLimitations of Escape AnalysisMonomorphic DispatchIntrinsicsOn-Stack ReplacementSafepoints RevisitedCore Library MethodsUpper Method Size Limit for InliningUpper Method Size Limit for CompilationSummary
Optimizing CollectionsOptimization Considerations for ListsArrayListLinkedListArrayList versus LinkedListOptimization Considerations for MapsHashMapTreeMapLack of MultiMapOptimization Considerations for SetsDomain ObjectsAvoid FinalizationWar Story: Forgetting to Clean UpWhy Not Use Finalization to Solve the Problem?try-with-resourcesMethod HandlesSummary
Introduction to ParallelismFundamental Java ConcurrencyUnderstanding the JMMBuilding Concurrency LibrariesUnsafeAtomics and CASLocks and SpinlocksSummary of Concurrent LibrariesLocks in java.util.concurrentRead/Write LocksSemaphoresConcurrent CollectionsLatches and BarriersExecutors and Task AbstractionIntroducing Asynchronous ExecutionSelecting an ExecutorServiceFork/JoinModern Java ConcurrencyStreams and Parallel StreamsLock-Free TechniquesActor-Based TechniquesSummary
Introduction to ProfilingSampling and Safepointing BiasExecution Profiling Tools for DevelopersVisualVM ProfilerJProfilerYourKitFlight Recorder and Mission ControlOperational ToolsModern ProfilersAllocation ProfilingHeap Dump AnalysishprofSummary
LoggingLogging MicrobenchmarksDesigning a Lower-Impact LoggerLow Latency Using Real Logic LibrariesAgronaSimple Binary EncodingAeronThe Design of AeronSummary
Small Performance Enhancements in Java 9Segmented Code CacheCompact StringsNew String ConcatenationC2 Compiler ImprovementsNew Version of G1Java 10 and Future VersionsNew Release ProcessJava 10Unsafe in Java 9 and BeyondVarHandles in Java 9Project Valhalla and Value TypesGraal and TruffleFuture Directions in BytecodeFuture Directions in ConcurrencyConclusion

Overview

Performance tuning is an experimental science, but that doesn’t mean engineers should resort to guesswork and folklore to get the job done. Yet that’s often the case. With this practical book, intermediate to advanced Java technologists working with complex technology stacks will learn how to tune Java applications for performance using a quantitative, verifiable approach.

Most resources on performance tend to discuss the theory and internals of Java virtual machines, but this book focuses on the practicalities of performance tuning by examining a wide range of aspects. There are no simple recipes, tips and tricks, or algorithms to learn. Performance tuning is a process of defining and determining desired outcomes. And it requires diligence.

Learn how Java principles and technology make the best use of modern hardware and operating systems
Explore several performance tests and common anti-patterns that can vex your team
Understand the pitfalls of measuring Java performance numbers and the drawbacks of microbenchmarking
Dive into JVM garbage collection logging, monitoring, tuning, and tools
Explore JIT compilation and Java language performance techniques
Learn performance aspects of the Java Collections API and get an overview of Java concurrency