Soumettre la recherche
Mettre en ligne
GC Tuning Confessions Of A Performance Engineer
•
46 j'aime
•
12,247 vues
Monica Beckwith
Suivre
GC Tuning Confessions
Lire moins
Lire la suite
Logiciels
Signaler
Partager
Signaler
Partager
1 sur 71
Recommandé
Garbage First Garbage Collector (G1 GC): Current and Future Adaptability and ...
Garbage First Garbage Collector (G1 GC): Current and Future Adaptability and ...
Monica Beckwith
The Performance Engineer's Guide To (OpenJDK) HotSpot Garbage Collection - Th...
The Performance Engineer's Guide To (OpenJDK) HotSpot Garbage Collection - Th...
Monica Beckwith
The Performance Engineer's Guide To HotSpot Just-in-Time Compilation
The Performance Engineer's Guide To HotSpot Just-in-Time Compilation
Monica Beckwith
Way Improved :) GC Tuning Confessions - presented at JavaOne2015
Way Improved :) GC Tuning Confessions - presented at JavaOne2015
Monica Beckwith
The Performance Engineer's Guide to Java (HotSpot) Virtual Machine
The Performance Engineer's Guide to Java (HotSpot) Virtual Machine
Monica Beckwith
GC Tuning Confessions Of A Performance Engineer - Improved :)
GC Tuning Confessions Of A Performance Engineer - Improved :)
Monica Beckwith
Game of Performance: A Song of JIT and GC
Game of Performance: A Song of JIT and GC
Monica Beckwith
Garbage First Garbage Collector: Where the Rubber Meets the Road!
Garbage First Garbage Collector: Where the Rubber Meets the Road!
Monica Beckwith
Recommandé
Garbage First Garbage Collector (G1 GC): Current and Future Adaptability and ...
Garbage First Garbage Collector (G1 GC): Current and Future Adaptability and ...
Monica Beckwith
The Performance Engineer's Guide To (OpenJDK) HotSpot Garbage Collection - Th...
The Performance Engineer's Guide To (OpenJDK) HotSpot Garbage Collection - Th...
Monica Beckwith
The Performance Engineer's Guide To HotSpot Just-in-Time Compilation
The Performance Engineer's Guide To HotSpot Just-in-Time Compilation
Monica Beckwith
Way Improved :) GC Tuning Confessions - presented at JavaOne2015
Way Improved :) GC Tuning Confessions - presented at JavaOne2015
Monica Beckwith
The Performance Engineer's Guide to Java (HotSpot) Virtual Machine
The Performance Engineer's Guide to Java (HotSpot) Virtual Machine
Monica Beckwith
GC Tuning Confessions Of A Performance Engineer - Improved :)
GC Tuning Confessions Of A Performance Engineer - Improved :)
Monica Beckwith
Game of Performance: A Song of JIT and GC
Game of Performance: A Song of JIT and GC
Monica Beckwith
Garbage First Garbage Collector: Where the Rubber Meets the Road!
Garbage First Garbage Collector: Where the Rubber Meets the Road!
Monica Beckwith
G1 collector and tuning and Cassandra
G1 collector and tuning and Cassandra
Chris Lohfink
OPENJDK: IN THE NEW AGE OF CONCURRENT GARBAGE COLLECTORS
OPENJDK: IN THE NEW AGE OF CONCURRENT GARBAGE COLLECTORS
Monica Beckwith
Java Performance Engineer's Survival Guide
Java Performance Engineer's Survival Guide
Monica Beckwith
JFokus Java 9 contended locking performance
JFokus Java 9 contended locking performance
Monica Beckwith
Garbage First Garbage Collector (G1 GC) - Migration to, Expectations and Adva...
Garbage First Garbage Collector (G1 GC) - Migration to, Expectations and Adva...
Monica Beckwith
Java Garbage Collectors – Moving to Java7 Garbage First (G1) Collector
Java Garbage Collectors – Moving to Java7 Garbage First (G1) Collector
Gurpreet Sachdeva
G1 Garbage Collector - Big Heaps and Low Pauses?
G1 Garbage Collector - Big Heaps and Low Pauses?
C2B2 Consulting
Enabling Java: Windows on Arm64 - A Success Story!
Enabling Java: Windows on Arm64 - A Success Story!
Monica Beckwith
Moving to g1 gc by Kirk Pepperdine.
Moving to g1 gc by Kirk Pepperdine.
J On The Beach
Storing Cassandra Metrics
Storing Cassandra Metrics
Chris Lohfink
-XX:+UseG1GC
-XX:+UseG1GC
Jakub Kubrynski
Hadoop world g1_gc_forh_base_v4
Hadoop world g1_gc_forh_base_v4
YanpingWang
G1GC
G1GC
koji lin
Tuning Java for Big Data
Tuning Java for Big Data
Scott Seighman
Performance tuning jvm
Performance tuning jvm
Prem Kuppumani
Moving to G1GC
Moving to G1GC
Kirk Pepperdine
JVM memory management & Diagnostics
JVM memory management & Diagnostics
Dhaval Shah
Jvm Performance Tunning
Jvm Performance Tunning
guest1f2740
Garbage First and you
Garbage First and you
Kai Koenig
Trouble with memory
Trouble with memory
Kirk Pepperdine
ZGC-SnowOne.pdf
ZGC-SnowOne.pdf
Monica Beckwith
JVM Performance Tuning
JVM Performance Tuning
Jeremy Leisy
Contenu connexe
Tendances
G1 collector and tuning and Cassandra
G1 collector and tuning and Cassandra
Chris Lohfink
OPENJDK: IN THE NEW AGE OF CONCURRENT GARBAGE COLLECTORS
OPENJDK: IN THE NEW AGE OF CONCURRENT GARBAGE COLLECTORS
Monica Beckwith
Java Performance Engineer's Survival Guide
Java Performance Engineer's Survival Guide
Monica Beckwith
JFokus Java 9 contended locking performance
JFokus Java 9 contended locking performance
Monica Beckwith
Garbage First Garbage Collector (G1 GC) - Migration to, Expectations and Adva...
Garbage First Garbage Collector (G1 GC) - Migration to, Expectations and Adva...
Monica Beckwith
Java Garbage Collectors – Moving to Java7 Garbage First (G1) Collector
Java Garbage Collectors – Moving to Java7 Garbage First (G1) Collector
Gurpreet Sachdeva
G1 Garbage Collector - Big Heaps and Low Pauses?
G1 Garbage Collector - Big Heaps and Low Pauses?
C2B2 Consulting
Enabling Java: Windows on Arm64 - A Success Story!
Enabling Java: Windows on Arm64 - A Success Story!
Monica Beckwith
Moving to g1 gc by Kirk Pepperdine.
Moving to g1 gc by Kirk Pepperdine.
J On The Beach
Storing Cassandra Metrics
Storing Cassandra Metrics
Chris Lohfink
-XX:+UseG1GC
-XX:+UseG1GC
Jakub Kubrynski
Hadoop world g1_gc_forh_base_v4
Hadoop world g1_gc_forh_base_v4
YanpingWang
G1GC
G1GC
koji lin
Tuning Java for Big Data
Tuning Java for Big Data
Scott Seighman
Performance tuning jvm
Performance tuning jvm
Prem Kuppumani
Moving to G1GC
Moving to G1GC
Kirk Pepperdine
JVM memory management & Diagnostics
JVM memory management & Diagnostics
Dhaval Shah
Jvm Performance Tunning
Jvm Performance Tunning
guest1f2740
Garbage First and you
Garbage First and you
Kai Koenig
Trouble with memory
Trouble with memory
Kirk Pepperdine
Tendances
(20)
G1 collector and tuning and Cassandra
G1 collector and tuning and Cassandra
OPENJDK: IN THE NEW AGE OF CONCURRENT GARBAGE COLLECTORS
OPENJDK: IN THE NEW AGE OF CONCURRENT GARBAGE COLLECTORS
Java Performance Engineer's Survival Guide
Java Performance Engineer's Survival Guide
JFokus Java 9 contended locking performance
JFokus Java 9 contended locking performance
Garbage First Garbage Collector (G1 GC) - Migration to, Expectations and Adva...
Garbage First Garbage Collector (G1 GC) - Migration to, Expectations and Adva...
Java Garbage Collectors – Moving to Java7 Garbage First (G1) Collector
Java Garbage Collectors – Moving to Java7 Garbage First (G1) Collector
G1 Garbage Collector - Big Heaps and Low Pauses?
G1 Garbage Collector - Big Heaps and Low Pauses?
Enabling Java: Windows on Arm64 - A Success Story!
Enabling Java: Windows on Arm64 - A Success Story!
Moving to g1 gc by Kirk Pepperdine.
Moving to g1 gc by Kirk Pepperdine.
Storing Cassandra Metrics
Storing Cassandra Metrics
-XX:+UseG1GC
-XX:+UseG1GC
Hadoop world g1_gc_forh_base_v4
Hadoop world g1_gc_forh_base_v4
G1GC
G1GC
Tuning Java for Big Data
Tuning Java for Big Data
Performance tuning jvm
Performance tuning jvm
Moving to G1GC
Moving to G1GC
JVM memory management & Diagnostics
JVM memory management & Diagnostics
Jvm Performance Tunning
Jvm Performance Tunning
Garbage First and you
Garbage First and you
Trouble with memory
Trouble with memory
Similaire à GC Tuning Confessions Of A Performance Engineer
ZGC-SnowOne.pdf
ZGC-SnowOne.pdf
Monica Beckwith
JVM Performance Tuning
JVM Performance Tuning
Jeremy Leisy
Choosing Right Garbage Collector to Increase Efficiency of Java Memory Usage
Choosing Right Garbage Collector to Increase Efficiency of Java Memory Usage
Jelastic Multi-Cloud PaaS
this-is-garbage-talk-2022.pptx
this-is-garbage-talk-2022.pptx
Tier1 app
millions-gc-jax-2022.pptx
millions-gc-jax-2022.pptx
Tier1 app
Java Performance Tuning
Java Performance Tuning
Ender Aydin Orak
Elastic JVM for Scalable Java EE Applications Running in Containers #Jakart...
Elastic JVM for Scalable Java EE Applications Running in Containers #Jakart...
Jelastic Multi-Cloud PaaS
Вячеслав Блинов «Java Garbage Collection: A Performance Impact»
Вячеслав Блинов «Java Garbage Collection: A Performance Impact»
Anna Shymchenko
淺談 Java GC 原理、調教和新發展
淺談 Java GC 原理、調教和新發展
Leon Chen
How to Become a Winner in the JVM Performance-Tuning Battle
How to Become a Winner in the JVM Performance-Tuning Battle
Capgemini
Taming GC Pauses for Humongous Java Heaps in Spark Graph Computing-(Eric Kacz...
Taming GC Pauses for Humongous Java Heaps in Spark Graph Computing-(Eric Kacz...
Spark Summit
State of Java Elasticity. Tuning Java Efficiency - GIDS.JAVA LIVE 2020
State of Java Elasticity. Tuning Java Efficiency - GIDS.JAVA LIVE 2020
Jelastic Multi-Cloud PaaS
Værktøjer udviklet på AAU til analyse af SCJ programmer
Værktøjer udviklet på AAU til analyse af SCJ programmer
InfinIT - Innovationsnetværket for it
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
HostedbyConfluent
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
HostedbyConfluent
Jvm problem diagnostics
Jvm problem diagnostics
Danijel Mitar
Software Profiling: Java Performance, Profiling and Flamegraphs
Software Profiling: Java Performance, Profiling and Flamegraphs
Isuru Perera
Java at Scale, Dallas JUG, October 2013
Java at Scale, Dallas JUG, October 2013
Azul Systems Inc.
Kubecon seattle 2018 workshop slides
Kubecon seattle 2018 workshop slides
Weaveworks
Secrets of Performance Tuning Java on Kubernetes
Secrets of Performance Tuning Java on Kubernetes
Bruno Borges
Similaire à GC Tuning Confessions Of A Performance Engineer
(20)
ZGC-SnowOne.pdf
ZGC-SnowOne.pdf
JVM Performance Tuning
JVM Performance Tuning
Choosing Right Garbage Collector to Increase Efficiency of Java Memory Usage
Choosing Right Garbage Collector to Increase Efficiency of Java Memory Usage
this-is-garbage-talk-2022.pptx
this-is-garbage-talk-2022.pptx
millions-gc-jax-2022.pptx
millions-gc-jax-2022.pptx
Java Performance Tuning
Java Performance Tuning
Elastic JVM for Scalable Java EE Applications Running in Containers #Jakart...
Elastic JVM for Scalable Java EE Applications Running in Containers #Jakart...
Вячеслав Блинов «Java Garbage Collection: A Performance Impact»
Вячеслав Блинов «Java Garbage Collection: A Performance Impact»
淺談 Java GC 原理、調教和新發展
淺談 Java GC 原理、調教和新發展
How to Become a Winner in the JVM Performance-Tuning Battle
How to Become a Winner in the JVM Performance-Tuning Battle
Taming GC Pauses for Humongous Java Heaps in Spark Graph Computing-(Eric Kacz...
Taming GC Pauses for Humongous Java Heaps in Spark Graph Computing-(Eric Kacz...
State of Java Elasticity. Tuning Java Efficiency - GIDS.JAVA LIVE 2020
State of Java Elasticity. Tuning Java Efficiency - GIDS.JAVA LIVE 2020
Værktøjer udviklet på AAU til analyse af SCJ programmer
Værktøjer udviklet på AAU til analyse af SCJ programmer
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
Better Kafka Performance Without Changing Any Code | Simon Ritter, Azul
Jvm problem diagnostics
Jvm problem diagnostics
Software Profiling: Java Performance, Profiling and Flamegraphs
Software Profiling: Java Performance, Profiling and Flamegraphs
Java at Scale, Dallas JUG, October 2013
Java at Scale, Dallas JUG, October 2013
Kubecon seattle 2018 workshop slides
Kubecon seattle 2018 workshop slides
Secrets of Performance Tuning Java on Kubernetes
Secrets of Performance Tuning Java on Kubernetes
Plus de Monica Beckwith
The ilities of software engineering.pptx
The ilities of software engineering.pptx
Monica Beckwith
A G1GC Saga-KCJUG.pptx
A G1GC Saga-KCJUG.pptx
Monica Beckwith
QCon London.pdf
QCon London.pdf
Monica Beckwith
Applying Concurrency Cookbook Recipes to SPEC JBB
Applying Concurrency Cookbook Recipes to SPEC JBB
Monica Beckwith
Intro to Garbage Collection
Intro to Garbage Collection
Monica Beckwith
OpenJDK Concurrent Collectors
OpenJDK Concurrent Collectors
Monica Beckwith
Java 9: The (G1) GC Awakens!
Java 9: The (G1) GC Awakens!
Monica Beckwith
Plus de Monica Beckwith
(7)
The ilities of software engineering.pptx
The ilities of software engineering.pptx
A G1GC Saga-KCJUG.pptx
A G1GC Saga-KCJUG.pptx
QCon London.pdf
QCon London.pdf
Applying Concurrency Cookbook Recipes to SPEC JBB
Applying Concurrency Cookbook Recipes to SPEC JBB
Intro to Garbage Collection
Intro to Garbage Collection
OpenJDK Concurrent Collectors
OpenJDK Concurrent Collectors
Java 9: The (G1) GC Awakens!
Java 9: The (G1) GC Awakens!
Dernier
Tech Tuesday Slides - Introduction to Project Management with OnePlan's Work ...
Tech Tuesday Slides - Introduction to Project Management with OnePlan's Work ...
OnePlan Solutions
The Role of IoT and Sensor Technology in Cargo Cloud Solutions.pptx
The Role of IoT and Sensor Technology in Cargo Cloud Solutions.pptx
RTS corp
eSoftTools IMAP Backup Software and migration tools
eSoftTools IMAP Backup Software and migration tools
osttopstonverter
Alfresco TTL#157 - Troubleshooting Made Easy: Deciphering Alfresco mTLS Confi...
Alfresco TTL#157 - Troubleshooting Made Easy: Deciphering Alfresco mTLS Confi...
Angel Borroy López
Powering Real-Time Decisions with Continuous Data Streams
Powering Real-Time Decisions with Continuous Data Streams
Safe Software
Sending Calendar Invites on SES and Calendarsnack.pdf
Sending Calendar Invites on SES and Calendarsnack.pdf
31events.com
Amazon Bedrock in Action - presentation of the Bedrock's capabilities
Amazon Bedrock in Action - presentation of the Bedrock's capabilities
KrzysztofKkol1
Tech Tuesday - Mastering Time Management Unlock the Power of OnePlan's Timesh...
Tech Tuesday - Mastering Time Management Unlock the Power of OnePlan's Timesh...
OnePlan Solutions
UI5ers live - Custom Controls wrapping 3rd-party libs.pptx
UI5ers live - Custom Controls wrapping 3rd-party libs.pptx
Andreas Kunz
VK Business Profile - provides IT solutions and Web Development
VK Business Profile - provides IT solutions and Web Development
vyaparkranti
Leveraging AI for Mobile App Testing on Real Devices | Applitools + Kobiton
Leveraging AI for Mobile App Testing on Real Devices | Applitools + Kobiton
Applitools
Keeping your build tool updated in a multi repository world
Keeping your build tool updated in a multi repository world
Roberto Pérez Alcolea
Introduction to Firebase Workshop Slides
Introduction to Firebase Workshop Slides
vaideheekore1
Machine Learning Software Engineering Patterns and Their Engineering
Machine Learning Software Engineering Patterns and Their Engineering
Hironori Washizaki
VictoriaMetrics Q1 Meet Up '24 - Community & News Update
VictoriaMetrics Q1 Meet Up '24 - Community & News Update
VictoriaMetrics
OpenChain Education Work Group Monthly Meeting - 2024-04-10 - Full Recording
OpenChain Education Work Group Monthly Meeting - 2024-04-10 - Full Recording
Shane Coughlan
Large Language Models for Test Case Evolution and Repair
Large Language Models for Test Case Evolution and Repair
Lionel Briand
Comparing Linux OS Image Update Models - EOSS 2024.pdf
Comparing Linux OS Image Update Models - EOSS 2024.pdf
Drew Moseley
Post Quantum Cryptography – The Impact on Identity
Post Quantum Cryptography – The Impact on Identity
team-WIBU
SoftTeco - Software Development Company Profile
SoftTeco - Software Development Company Profile
akrivarotava
Dernier
(20)
Tech Tuesday Slides - Introduction to Project Management with OnePlan's Work ...
Tech Tuesday Slides - Introduction to Project Management with OnePlan's Work ...
The Role of IoT and Sensor Technology in Cargo Cloud Solutions.pptx
The Role of IoT and Sensor Technology in Cargo Cloud Solutions.pptx
eSoftTools IMAP Backup Software and migration tools
eSoftTools IMAP Backup Software and migration tools
Alfresco TTL#157 - Troubleshooting Made Easy: Deciphering Alfresco mTLS Confi...
Alfresco TTL#157 - Troubleshooting Made Easy: Deciphering Alfresco mTLS Confi...
Powering Real-Time Decisions with Continuous Data Streams
Powering Real-Time Decisions with Continuous Data Streams
Sending Calendar Invites on SES and Calendarsnack.pdf
Sending Calendar Invites on SES and Calendarsnack.pdf
Amazon Bedrock in Action - presentation of the Bedrock's capabilities
Amazon Bedrock in Action - presentation of the Bedrock's capabilities
Tech Tuesday - Mastering Time Management Unlock the Power of OnePlan's Timesh...
Tech Tuesday - Mastering Time Management Unlock the Power of OnePlan's Timesh...
UI5ers live - Custom Controls wrapping 3rd-party libs.pptx
UI5ers live - Custom Controls wrapping 3rd-party libs.pptx
VK Business Profile - provides IT solutions and Web Development
VK Business Profile - provides IT solutions and Web Development
Leveraging AI for Mobile App Testing on Real Devices | Applitools + Kobiton
Leveraging AI for Mobile App Testing on Real Devices | Applitools + Kobiton
Keeping your build tool updated in a multi repository world
Keeping your build tool updated in a multi repository world
Introduction to Firebase Workshop Slides
Introduction to Firebase Workshop Slides
Machine Learning Software Engineering Patterns and Their Engineering
Machine Learning Software Engineering Patterns and Their Engineering
VictoriaMetrics Q1 Meet Up '24 - Community & News Update
VictoriaMetrics Q1 Meet Up '24 - Community & News Update
OpenChain Education Work Group Monthly Meeting - 2024-04-10 - Full Recording
OpenChain Education Work Group Monthly Meeting - 2024-04-10 - Full Recording
Large Language Models for Test Case Evolution and Repair
Large Language Models for Test Case Evolution and Repair
Comparing Linux OS Image Update Models - EOSS 2024.pdf
Comparing Linux OS Image Update Models - EOSS 2024.pdf
Post Quantum Cryptography – The Impact on Identity
Post Quantum Cryptography – The Impact on Identity
SoftTeco - Software Development Company Profile
SoftTeco - Software Development Company Profile
GC Tuning Confessions Of A Performance Engineer
1.
GC Tuning Confessions Of
A Performance Engineer Monica Beckwith monica@codekaram.com @mon_beck www.linkedin.com/in/monicabeckwith Philly Emerging Tech Conference 2015 April 8th, 2015
2.
©2015 CodeKaram About Me •
JVM/GC Performance Engineer/Consultant • Worked at AMD, Sun, Oracle • Worked with HotSpot JVM • Worked with Parallel(Old) GC, G1 GC and CMS GC 2
3.
©2015 CodeKaram About Today’s
Talk • A little bit about Performance Engineering • Insight into Garbage Collectors • Introduction to a few main GC Algorithms in OpenJDK HotSpot (ParallelOld GC, CMS GC, and G1 GC) • Summary • GC Tunables 3
4.
©2015 CodeKaram Performance Engineering •
A performance engineer helps ensure that the system is designed + implemented to meet the performance requirements. • The performance requirements could include the service level agreements (SLAs) for throughput, latency and other response time related metrics - also known as non-functional requirements. • E.g. Response time (RT) metrics - Average (RT), max or worst-case RT, 99th percentile RT… • Let’s talk more about these RTs… 4
5.
©2015 CodeKaram Performance Engineering
- Response Time Metrics Average (ms) Minimum (ms) System1 307.741 7.622 System2 320.778 7.258 System3 321.483 6.432 System4 323.143 7.353 5
6.
©2015 CodeKaram Performance Engineering
- Response Time Metrics Average (ms) Number of GCs System1 307.741 37353 System2 320.778 34920 System3 321.483 36270 System4 323.143 40636 6
7.
©2015 CodeKaram Performance Engineering
- Response Time Metrics Average (ms) Maximum (ms) System1 307.741 3131.331 System2 320.778 2744.588 System3 321.483 1681.308 System4 323.143 20699.505 7
8.
©2015 CodeKaram Performance Engineering
- Response Time Metrics Average (ms) Maximum (ms) System1 307.741 3131.331 System2 320.778 2744.588 System3 321.483 1681.308 System4 323.143 20699.505 5 full GCs and 10 evacuation failures 8
9.
©2015 CodeKaram Performance Engineering •
Monitoring, analysis and tuning are a big part of performance engineering. • Monitoring utilization - CPU, IO, Memory bandwidth, Java heap, … • Analyzing utilization and time spent - GC logs, CPU, memory and application logs • Profiling - Application, System, Memory - Java Heap. • Java/JVM performance engineering includes the study, analysis and tuning of the Just-in-time (JIT) compiler, the Garbage Collector (GC) and many a times tuning related to the Java Development Kit (JDK). 9
10.
©2015 CodeKaram Insight Into
Garbage Collectors (GCs) • A GC is an automatic memory management unit. • An ideal GC is the one that requires minimum footprint (concurrent CPU or native memory), and provides maximum throughput while minimizing predictable latency. • Fun Fact - In reality you will have to tradeoff one (footprint or latency or throughput) in lieu of the others. A healthy performance engineering exercise can help you meet or exceed your goals. • Fun Fact - GC can NOT eliminate your memory leaks! • Fun Fact - GC (and heap dump) can provide an insight into your application. 10
11.
©2015 CodeKaram GC Algorithms
in OpenJDK HotSpot - The Tradeoff • Throughput and latency are the two main drivers towards refinement of GC algorithms. • Fun Fact - Most OpenJDK HotSpot users would like to increase their (Java) heap space but they fear full garbage collections. 11
12.
©2015 CodeKaram GC Algorithms
in OpenJDK HotSpot - Throughput Maximizer • Throughput has driven us to parallelization of GC worker threads: • Parallel Collection Threads • Parallel Concurrent Marking Threads • Throughput has driven us to generational GCs • Most objects die young. • Fast path allocation into “young” generation. • Age and then promote (fast path again) to “old” generation • Fun Fact: All GCs in OpenJDK HotSpot are generational. 12
13.
©2015 CodeKaram GC Algorithms
in OpenJDK HotSpot - Latency Sensitive • Latency has driven algorithms to no compaction or partial compaction • Time is of essence, no need to fully compact if free space is still available! • Latency has driven algorithms towards concurrency - i.e. running with the application threads. • Mostly concurrent mark and sweep (CMS) and concurrent marking in G1. • Fun Fact: All GCs in OpenJDK HotSpot fallback to a fully compacting stop-the-world garbage collection called the “full” GC. • Tuning can help avoid or postpone full GCs in many cases. 13
14.
©2015 CodeKaram The Throughput
Collector • ParallelOld is the throughput collector in OpenJDK HotSpot. • But, first, what is throughput? • Throughput is the percentage of time NOT spent in GC :) • The throughput goal for ParallelOld Collector is 99%. • That is, all the GC pauses that happen during the life of the application should account to 1% of the run time. • How does ParallelOld try to achieve its throughput goal? 14
15.
©2015 CodeKaram The Throughput
Collector - Java Heap Eden S0 S1 Old Generation Young Generation Allocations Survivors Tenured 15
16.
©2015 CodeKaram • An
allocation failure results in a stop-the-world young collection. • The young generation is collected in its entirety i.e. all objects (dead or alive) are emptied from the eden and S0 spaces. • After the young collection is complete, the surviving objects (objects that are live) are moved into S1. • Objects are aged in the survivor space until ready for promotion • When the age threshold is met, objects are promoted into the old generation. • Allocations and promotions are both fast tracked (lock-free) by using Thread/Promotion Local Allocation Buffers (TLAB/PLAB) The Throughput Collector - Young Collection 16
17.
©2015 CodeKaram Old Generation Free
Space Occupied Space The Throughput Collector - Contiguous Old Generation 17
18.
©2015 CodeKaram Old Generation To-be
Promoted Object 1 The Throughput Collector - Contiguous Old Generation 18
19.
©2015 CodeKaram Old Generation Free
Space Occupied Space The Throughput Collector - Contiguous Old Generation 19
20.
©2015 CodeKaram Old Generation To-be
Promoted Object 2 The Throughput Collector - Contiguous Old Generation 20
21.
©2015 CodeKaram Old Generation Free
Space Occupied Space The Throughput Collector - Contiguous Old Generation 21
22.
©2015 CodeKaram Old Generation To-be
Promoted Object 3 The Throughput Collector - Contiguous Old Generation 22
23.
©2015 CodeKaram • When
the old generation is full, i.e. when the old generation can’t accept any promotions, it is time to start a full compaction collection. • The ParallelOld collector utilized parallel stop-the-world GC threads to help compact the entire heap. • The ParallelOld collector achieves compaction via copying and moving live data. • After the end of the full collection, only the compacted live data set of the java application occupies the old generation. The rest of the space in the old generation is free for future promotions. The Throughput Collector - Contiguous Old Generation 23
24.
©2015 CodeKaram Young GC Threads Java Application Threads Old GC Threads Java Application Threads Java Application Threads The
Throughput Collector 24
25.
©2015 CodeKaram The CMS
Collector • Mostly Concurrent Mark & Sweep collector is geared towards latency sensitive applications. • The young collections are pretty similar to what you would observe with ParallelOld GC’s young collection • The main difference is in the old generation collection. • An occupancy threshold determines when the mostly concurrent mark and sweep cycle can start. • The sweep cycle just reclaims the dead objects and updates a list of free spaces. Thus, the CMS collector is not a compacting collector. • This can lead to fragmentation which can lead to promotion failures which eventually lead to a fallback single-threaded fully compacting collection. 25
26.
©2015 CodeKaram The CMS
Collector • The CMS collector has multiple stop-the-world and concurrent phases - • After the old generation occupied space reaches or crosses the occupancy threshold, a CMS cycle is initiated. • Initially, parallel stop-the-world CMS threads help find the GC roots to start a live object graph. • Concurrent CMS threads help in marking further live objects (objects that are reachable by the roots). • Parallel stop-the-world CMS threads help in making sure that the final changes to the graph are duly noted. • Concurrent sweep phase helps in collecting all dead objects and updating the free list to track the reclaimed spaces. 26
27.
©2015 CodeKaram The CMS
Collector Young GC Threads Java Application Threads CMS Initial Mark Threads Java Application Threads Java Application Threads Young GC Threads CMS Remark Threads Java Application Threads Java Application Threads Concurrent CMS Threads Concurrent CMS Threads Concurrent CMS Threads 27
28.
©2015 CodeKaram The CMS
Collector - Contiguous Old Generation 28 Old Generation Free List Free Space Occupied Space
29.
©2015 CodeKaram Old Generation To-be
Promoted Object 1 The CMS Collector - Contiguous Old Generation 29
30.
©2015 CodeKaram Old Generation Free
List Free Space Occupied Space The CMS Collector - Contiguous Old Generation 30
31.
©2015 CodeKaram Old Generation To-be
Promoted Object 2 The CMS Collector - Contiguous Old Generation 31
32.
©2015 CodeKaram Old Generation Free
List Free Space Occupied Space The CMS Collector - Contiguous Old Generation 32
33.
©2015 CodeKaram X Old Generation To-be
Promoted Object 3 The CMS Collector - Contiguous Old Generation 33
34.
©2015 CodeKaram X Old Generation To-be
Promoted Object 3 The CMS Collector - Contiguous Old Generation 34
35.
©2015 CodeKaram X Old Generation To-be
Promoted Object 3 The CMS Collector - Contiguous Old Generation 35
36.
©2015 CodeKaram X Old Generation To-be
Promoted Object 3 The CMS Collector - Contiguous Old Generation 36
37.
©2015 CodeKaram Old Generation To-be
Promoted Object 3 ?? The CMS Collector - Contiguous Old Generation 37
38.
©2015 CodeKaram Old Generation To-be
Promoted Object 3 Promotion Failure!! The CMS Collector - Contiguous Old Generation 38
39.
©2015 CodeKaram • When
CMS just can’t keep up with the promotion rate • Your old generation is getting filled before a concurrent cycle can free up space and complete. • Causes - marking threshold is too high, heap too small, or high application mutation rate 39 The CMS Collector - Concurrent Mode Failures
40.
©2015 CodeKaram The Garbage
First Collector • Garbage First (G1) is the newest collector in OpenJDK HotSpot. • G1 GC is supposed to be a long term replacement of CMS GC. • G1 GC introduces the concept of regionalized heap. • The idea is to set a pause time goal which will be used as a hint by the ergonomics and prediction logic and G1 GC will try to adjust its generations and number of regions per collections so as to best meet its pause time goal. • G1 has multi-staged concurrent marking phase as well as stop the world collections for young and old generations 40
41.
©2015 CodeKaram The Garbage
First Collector • Throughput is a secondary consideration for G1. Hence the throughput goal is to have about 10% of the total time be spent in GC activities. • G1 GC aims to provide more predictable pause times. • The G1 GC regionalized heap ensures that the generations don’t have to be contiguous (as they were in ParallelOld and CMS GCs). • With regions, G1 just needs to know the limits for each generation and the pause time goal. • G1 will add regions as needed from the list of free regions. • Occupied regions can be eden, survivor, old and humongous (more on this later). 41
42.
©2015 CodeKaram The Garbage
First Collector - Regionalized Heap Eden Old Old Eden Old Survivor Humongous 42
43.
©2015 CodeKaram • For
G1 GC, the young generation consists of eden regions and survivor regions. • A region is a unit of collection • The region size is determined at JVM startup (and can be changed if needed) • Most allocations are fast path allocations into the eden regions. • The survivor regions help in aging the objects which are eventually promoted into the old generation after the age threshold is crossed • The young generation is resized after every collection based on the time it took the current collection, the pause time goal and other factors considered by the prediction logic. The Garbage First Collector - The Young Generation 43
44.
©2015 CodeKaram The Garbage
First Collector Eden Old Old Eden Old Surv ivor E.g.: Current heap configuration - 44
45.
©2015 CodeKaram The Garbage
First Collector Eden Old Old Eden Old Surv ivor E.g.: During a young collection - 45
46.
©2015 CodeKaram The Garbage
First Collector Old Old Old E.g.: After a young collection - Sur vivor Old 46
47.
©2015 CodeKaram • For
G1 GC, the old generation collection consists of all of the young regions and a few candidate old regions. Such a collection is called a mixed collection. • The GC ergonomics selects the candidate old regions based on certain thresholds and calculations that help determine the regions with the most reclaimable space and at the same time giving consideration to the pause time goal. • Based on the total number of candidate old regions, the pause time goal and other collection related thresholds, you could see more that one mixed collection pause. Thus the old generation is collected incrementally. The Garbage First Collector - The Old Generation 47
48.
©2015 CodeKaram • In-order
to start a mixed collection, the heap occupancy threshold (also known as the marking threshold) must be crossed. • When the marking threshold is crossed, G1 GC initiates a concurrent marking cycle. The concurrent cycle is mostly concurrent, except for when marking the roots and when remarking to make sure all the mutations are captured appropriately in the object graph. • During the concurrent marking phase, any garbage-filled regions are reclaimed as a part of the cleanup phase. The Garbage First Collector - The Marking Threshold 48
49.
©2015 CodeKaram The Garbage
First Collector Old Old Old E.g.: Current heap configuration - Sur vivor Old Eden Eden Old 49
50.
©2015 CodeKaram Old Old Old E.g.:
Reclamation of a garbage-filled region during a concurrent cleanup phase - Sur vivor Old Eden Eden Old The Garbage First Collector 50
51.
©2015 CodeKaram Old Old Old E.g.:
Current heap configuration - Sur vivor Old Eden Eden The Garbage First Collector 51
52.
©2015 CodeKaram Old Old Old E.g.:
During a mixed collection - Sur vivor Old Eden Eden The Garbage First Collector 52
53.
©2015 CodeKaram Old E.g.: After
a mixed collection - O ld Surv ivor Old The Garbage First Collector 53
54.
©2015 CodeKaram • Objects
>= 50% of G1 region size == Humongous Objects • These objects are allocated directly into the old generation into Humongous Regions • If a humongous object is bigger than a G1 region, then the humongous regions that contain it have to be contiguous. • Ideally, humongous objects are few in number and are short lived. • Can cause evacuation failures if application allocates a lot of long-lived humongous objects, since humongous regions add to the old generation occupancy. The Garbage First Collector - Humongous Objects 54
55.
©2015 CodeKaram • Evacuation
failures or promotion failures or to- space overflow or exhausted all refer to the same notion. • When there are no more regions available for survivors or tenured objects, G1 GC encounters an evacuation failure. • An evacuation failure is expensive and the usual pattern is that if you see a couple of evacuation failures; full GC will soon follow. The Garbage First Collector - Evacuation Failures 55
56.
©2015 CodeKaram • Make
sure your command line is not overloaded - Avoid over- tuning! • Try increasing your heap size. • Check if humongous allocations are the cause of your problem. • Adjust the marking threshold • Try increasing your concurrent threads to help reduce the time for concurrent marking phase. • If survivor objects are the issue, try increase the G1ReservePercent The Garbage First Collector - Avoiding Evacuation Failures 56
57.
©2015 CodeKaram What have
we learned so far? - Young Generation & Collections • Young generation is always collected in its entirety. • All 3 GCs discussed earlier follow similar mechanism for young collection. • The young collections achieve reclamation via compaction and copying of live objects. • There are a lot of options for sizing the Eden and Survivor space optimally and many GCs also have adaptive sizing and ergonomics for young generation collections. 57
58.
©2015 CodeKaram Most Allocations
Eden Space Eden Full? Start Young Collection: Keep Allocating :) Objects Aged in Survivor Space? Promote: Keep Aging :) Fast Path Promotions Fast Path Old Generation CMS promotes to fitting free space out of a free list. What have we learned so far? - Young Generation & Collections 58
59.
©2015 CodeKaram All 3
GCs vary in the way they collect the old generation: • For ParallelOld GC, the old generation is reclaimed and compacted in its entirety • Luckily, the compaction cost is distributed amongst parallel garbage collector worker threads. • Unluckily, the compaction cost depends a lot on the size of the live data set since at every compaction, the GC is moving live data around. • No tuning options other than the generation size adjustment and age threshold for promotion. 59 What have we learned so far? - Old Generation & Collections
60.
©2015 CodeKaram • For
CMS GC, the old generation is (mostly) concurrently marked and swept. Thus the reclamation of dead objects happen in place and the space is added to a free list of spaces. • The marking threshold can be tuned adaptively and manually as well. • Luckily, CMS GC doesn’t do compaction, hence reclamation is fast. • Unluckily, a long running Java application with CMS GC is prone to fragmentation which will eventually result in promotion failures which can eventually lead to full compacting garbage collection and sometimes even concurrent mode failures. • Full compacting GCs are singled threaded in CMS GC. What have we learned so far? - Old Generation & Collections 60
61.
©2015 CodeKaram • For
G1 GC, the old generation regions are (mostly) concurrently marked and an incremental compacting collection helps with optimizing the old generation collection. • Luckily, fragmentation is not “untunable” in G1 GC as it is in CMS GC. • Unluckily, sometimes, you may still encounter promotion/evacuation failures when G1 GC runs out of regions to copy live objects. Such an evacuation failure is expensive and can eventually lead to a full compacting GC. • Full compacting GCs are singled threaded in G1 GC. • Appropriate tuning of the old generation space and collection can help avoid evacuation failures and hence keep full GCs at bay. • G1 GC has multiple tuning options so that the GC can be adapted to your application needs. 61 What have we learned so far? - Old Generation & Collections
62.
©2015 CodeKaram GC Tunables
- The Throughput Collector Goal: Only promote objects after you have hazed them appropriately aged Tunables: Everything related to aging objects and generation sizing - NewRatio, (Max)NewSize, SurvivorRatio, (Max)TenuringThreshold 62
63.
©2015 CodeKaram GC Tunables
- The Throughput Collector Things to remember - • Applications with steady behavior rarely need AdaptiveSizePolicy to be enabled. • Overflow gets promoted into the old generation • Provide larger survivor spaces for long-lived transient data. • In most cases, young generation sizing has the most effect on throughput • Size the young generation to maintain the GC overhead to less than 5%. 63
64.
©2015 CodeKaram GC Tunables
- The CMS Collector Goal: Only promote objects after you have hazed them appropriately aged Tunables: Everything related to aging objects and young generation sizing still applies here. The concurrent thread counts and marking threshold are addition tunables for CMS 64
65.
©2015 CodeKaram GC Tunables
- The CMS Collector Things to remember - • Premature promotions are very expensive in CMS and could lead to fragmentation • You can reduce the CMS cycle duration by adding more concurrent threads: ConcGCThreads. • Remember that this will increase the concurrent overhead. • You can manually tune the marking threshold (adaptive by default) • CMSInitiatingOccupancyFraction & UseCMSInitiatingOccupancyOnly will help fix the marking threshold. • Note: The threshold is expressed as a percentage of the old generation occupancy 65
66.
©2015 CodeKaram GC Tunables
- The G1 Collector Goal: Get the ergonomics to work for you and know the defaults Tunables: • Pause time goal, heap size, max and min nursery, concurrent and parallel threads • The marking threshold, number of mixed GCs after marking, liveness threshold for the old regions, garbage toleration threshold, max old regions to be collected per mixed collection 66
67.
©2015 CodeKaram GC Tunables
- The G1 Collector Things to remember - • Know your defaults! • Understand your G1HeapRegionSize - It could be any factor of two from 1MB to 32MB. G1 strives for 2048 regions. • Fixing the nursery size (using Xmn) will meddle with the GC ergonomics. • Don’t set really aggressive pause time goals - this will increase the GC overhead. • Spend time taming your mixed GCs - mixed GCs are incremental old generation collections 67
68.
©2015 CodeKaram GC Tunables
- The G1 Collector Things to remember - • Taming mixed GCs: • Adjust the marking cycle according to you live data set. • Adjust you liveness threshold - this is the live occupancy threshold per region. Any region with liveness beyond this threshold will not be included in a mixed collection. • Adjust your garbage toleration threshold - helps G1 not get too aggressive with mixed collections • Distribute mixed GC pauses over a number of mixed collections - adjust your mixed GC count target and change your max old region threshold percent so that you can limit the old regions per collection 68
69.
©2015 CodeKaram Further Reading 69
70.
©2015 CodeKaram Further Reading 70
71.
©2015 CodeKaram Further Reading •
Jon Masa’s blog: https://blogs.oracle.com/ jonthecollector/entry/our_collectors • A few of my articles on InfoQ: http:// www.infoq.com/author/Monica-Beckwith • Presentations: http://www.slideshare.net/ MonicaBeckwith • Mail archives on hotspot-gc-use@openjdk.java.net & hotspot-gc-dev@openjdk.java.net 71