This document discusses JVM tuning and garbage collection. It provides an overview of key JVM flags and garbage collection strategies. It also describes the IDI test case which includes a swing desktop client, wicket web application, mobile web app in development, multiple application servers and frameworks. Hardware choices like CPU architecture and memory are important considerations for performance. The ideal configuration depends on the application's needs in terms of computational intensity and concurrency.

1. HotSpot JVM TuningBy: Gilad GaronTechnicalMeeting!

2. -Xmx2gIt’s more than just:What is JVM Tuning?It’s a combination of:Hardware selection

3. Software architectureBut most of all, JVM Tuning is about:Understanding how your application works!

4. The IDI Test CaseChoosing the right hardwareGarbage CollectorsJVM TuningAgendaJVM Tuning

5. Swing desktop client applicationWicket web applicationMobile web application (In development)Multiple AS:JBoss Clusters, Apache Cluster, Tomcat & TerracottaMultiple frameworks: Spring, Guice, Wicket, Hibernate, Terracotta DSO, XStream, EJB 3.0, JMS, CGLib & Proprietary frameworks Technical InformationThe IDI Test Case

6. The IDI Test Case

7. 3000 +/- EJB Invocations per minute100+ concurrent web sessionsInternal Print servers clients Internal Mail servers clientsInternal Webservices clientsJBoss EAR cluster user load informationThe IDI Test Case

8. The IDI Test CaseChoosing the right hardwareGarbage CollectorsJVM TuningAgendaJVM Tuning

9. How many software engineers does it take to change a light bulb?Hardware choices affects code performanceHardware choices affects OS offerings (Do you really want Windows?)Hardware isn’t easily replaced, it costs a lot.Choosing the right hardwareNone. It's a hardware problem.

10. Which CPU architecture is right for you?Higher clock speed? (what about Moore’s law?)Higher core count?Number of hardware threads?Choosing the right hardwareRemember:Not all CPU’s are created equally

11. Highly computational apps relay on clock speed

12. Highly concurrent / multi user app relay on number of threads

13. Hardware benchmarking helps a lot, but it’s difficult to arrangeIntel Xeon / Itanium?AMD Operton?Sun SPARC?IBM Power?What should be your opinion?Choosing the right hardwareThe winner is….No one will ever ask for your opinion (almost)

14. IDI Hardware historyChoosing the right hardware @ IDI

15. The IDI Test CaseChoosing the right hardwareGarbage CollectorsJVM TuningAgendaJVM Tuning

16. Remember this?Garbage CollectorsGarbage Collection means that don’t have to anymore

17. When the JVM starts, the host OS assigns a dedicated memory space to that VMThe VM allocates memory to the application within that dedicated memory spaceThe VM frees memory automatically via garbage collectorsGarbage Collection is an expensive algorithmJVM Memory HandlingGarbage Collectors

18. The memory space is divided into three separated memory spaces:Eden SpaceWhere objects are bornSurvivor SpacesWhere objects matureTenure SpaceWhere objects grow old and dieGenerational Memory SpaceGarbage Collectors

19. HotSpot VM‘s Generational Memory SpaceGarbage CollectorsEden SpaceSurvivor Space 2Survivor Space 1Tenure (Old) GenerationPermanent Generation

20. ConcurrencyConcurrency describes its ability to do work while application threads are still runningParallelismParallelismdescribes the collector's ability to perform its work across multiple threads of execution.Hence, a collector can be parallel but not concurrent, concurrent but not parallel, or both parallel and concurrent.Technical Terms:Garbage Collectors

21. Copy CollectorA collector that copies live object from one space to anotherLow Pause CollectorA collector that works along side the application threadsThroughput CollectorA collector that doesn’t add to the serial execution time of the applicationTechnical Terms:Garbage Collectors

22. Memory FragmentationThe amount of unusable free memory (small holes that object can’t be fitted in)Stop the world:The application threads are stopped during the garbage collectionMinor CollectionA garbage collection of the young generationMajor CollectionA garbage collection of the old generationTechnical Terms, Continued:Garbage Collectors

23. Types of GC Strategies (Generational):Garbage Collectors

24. Garbage Collectors“Parallel New” StrategyYoung generation collection is done in parallel (CMS Version)

25. Old generation collection is done in serial“Parallel” StrategyYoung generation collection is done in parallel

26. Old generation collection is done in serial“Parallel Old” StrategyYoung generation collection is done in parallelOld generation collection is done mostly in parallel

27. Low pause collector (short Stop the World pauses)Uses Parallel New for new generation GCUses CMS for old generation GCNon compacting old generation collectorStarts before the memory space is full (68%)Requires a larger heapConcurrent Mark-Sweep StrategyGarbage Collectors

28. Serial Collector (New Generation Space)Garbage Collectors

29. Serial GC Strategy - Young GenerationGarbage CollectorsNo more room! Promoting to Tenure spaceEden memory spaceGarbage Collection!Resuming ApplicationSurvived!Earned it’s place inTenure spaceSurvivor space (To)Survivor space (From)Tenure memory space

30. Serial Collector (Old Generation Space)Garbage Collectors

31. Concurrent Mark-SweepGarbage Collectors

32. CMS - Old GenerationGarbage CollectorsTenure memory spaceGarbage Collection!Resuming ApplicationStop The World!Phase:Initial MarkConcurrent MarkRemarkConcurrent SweepConcurrent Pre-Clean

33. High probability soft real time collectorOne memory space for the young and tenure generationEntire memory space is divided into equally sized cardsCards are managed by regions and tracked by “Remembered Sets” for changesA global card table track all the cardsGarbage First (G1) What’s the big deal?Garbage Collectors

34. Garbage First (G1) - Heap StructureGarbage CollectorsCard(512 Bytes)Region ACard(512 Bytes)Region BGlobal Card TableCard(512 Bytes)Region NRemembered Set (Per Region)

35. Garbage First - Minor CollectionGarbage CollectorsSimilar to Parallel NewCompacts young regions Also compacts old regions if there’s timeGarbage First - Major CollectionReclaims empty regions at once

36. Marks low lived ratio regions for collection in the minor collection phaseIDI GC Problem Related HistoryGarbage Collectors @ IDI

37. IDI Garbage Collection Strategy:Garbage Collectors @ IDI

38. The IDI Test CaseChoosing the right hardwareGarbage CollectorsJVM TuningAgendaJVM Tuning

39. This is JVM Tuning:JVM Tuning

40. -Xmx<size><units>-Xms<size><units>-d64-server-XX:MaxPermSize=<size><units>-XX:-HeapDumpOnOutOfMemoryErrorThe must know flagsJVM Tuning

41. -Xloggc:<file_name>-XX:+ PrintHeapAtGC-XX:+ PrintGCDateStamps-XX:+ PrintGCDetails-XX:+ PrintGCApplicationStoppedTime-XX:+ PrintGCApplicationConcurrentTimeGC LogsJVM Tuning

42. -XX:+ UseSerialGC-XX:+ UseParNewGC-XX:+UseConcMarkSweepGC -XX:+ UseParallelGC-XX:+UseParallelOldGC-XX:+UseG1GC (JDK 1.7)-XX:+UnlockExperimentalVMOptions(for G1 JDK 1.6_14) Garbage Collector StrategiesJVM Tuning

43. Tuning GCJVM Tuning

44. Tuning GCJVM Tuning

45. Tuning GCJVM Tuning

46. Tuning GCJVM TuningAnd many many more…

47. Java SE 6 HotSpot[tm] Virtual Machine Garbage Collection TuningMemory Management WhitepaperGarbage-First Garbage Collection WhitepaperSetting Java GC Parameters For Large HeapsJava Tuning White PaperJava HotSpot VM OptionsThe most complete list of -XX options for Java 6 JVMG1: Java's Garbage First Garbage Collector`References

48. The Garbage-First Garbage Collector (JavaOne 2008)JVM Internals - Garbage Collection & Runtime OptimizationsGarbage Collection in the Java HotSpot Virtual MachineReferences