Introduction to Garbage Collection

Introduction to Garbage Collection Algorithms. Use cases in JVM.

1. Garbage Collection

2. In computer science, garbage collection (GC) is a
form of Automatic Memory Management.



The garbage collector attempts to reclaim the
memory occupied by objects that are no longer in
use by the program.

Garbage collection was invented by John McCarthy
around 1959 to abstract away manual memory
management in Lisp.

3. Static vs Dynamic memory allocation



Creating an array of ten integers in C (static)
int array[10];
Creating an array of integers in C (dynamic)
int * array = malloc(N * sizeof(int));

free(array);
In C++
Foo* fooPtr = new Foo()
delete fooPtr

4. Languages without GC

C, C++*, D*, Objective-C*,
Rust*

Languages

Languages wit GC

Java, Go, PHP, Python, Scala*,
Haskell, ….

5. Garbage Collection Algorithms

Reference Counting Mark and Sweep
Copy Collection
Generational Collection

6. Reference Counting


Keep an extra integer (“reference count”) to every
heap-allocated data structure.

●With a new reference ++refCount.
●When a reference disappears --refCount.
●If refCount == 0, then reclaim the storage
+Easy to implement
+Real Time cleanup
-Additional storage
-Speed: incr, decr
-Cycles not cleaned
C++, Objective-C, Rust,
PHP, Python….

7. Tracing Collectors vs Reference Counting

8. Mark and Sweep

Mark Phase: traverses all objects,
starting with roots, and marks every
object found as alive.
Sweep Phase: traverses all
objects, reclaim the storage of
unmarked objects.
+Handles cycles
+Easy to implement
-Stops the world**
-Scans the entire heap

9. Copy Collection

+Handles cycles
+Automatic Compaction
-Stops the world**
-Changes addresses
-2nd half is unused

10. Generational Collection

●Most objects die young
●Newer objects usually point to older objects
+Same as Copy (unidirectional)
+Less Objects to Copy
+Frequency of collection
+Faster
-Stops the world**
-Complex to implement

11. Java Stack vs Heap

12. Pre and Post Java 8 Heap

13. Java Garbage Collectors

●Serial Collector - -XX:+UseSerialGC
●Parallel Collector - -XX:+UseParallelGC
●Concurrent Mark & Sweep Collector - -XX:+UseConcMarkSweepGC
●Garbage First (G1) Collector - -XX:+UseG1GC

14. Serial GC

●Stops the world
●Uses single thread
●Designed for single CPU small Heap apps
●Do not use it!
●Young Gen - Copy Collector
●Old Gen - Mark and Sweep

15. Parallel GC (The Default Collector)

●Stops the world
●Designed to work with multiple CPUs
●Uses multiple threads
●Expect high latencies when GC runs (-XX:MaxGCPauseMillis=<N>)
●Young Gen - Copy Collector
●Old Gen - Mark and Sweep

16. Concurrent Mark and Sweep (CMS) GC

●Stops the world (relatively short pauses)
●Designed to work with multiple CPUs
●Uses multiple threads
●Good for low latency apps
●Young Gen - Copy Collector
●Old Gen - Concurrent Mark-Sweep

17. Garbage First (G1) Collector

●Heap is split into (typically 2048) smaller regions
●Avoids collecting the entire heap at once, instead collects incrementally
○The regions that contain the most garbage are collected first
●Soft real-time garbage collector (Predictable | Configurable STW)
●Compaction is relatively easy
●Uses multiple threads and is good for > 6G heap sizes. Java 9 Default GC.
●Young Gen - Copy Collector
●Old Gen - Concurrent Mark-Sweep

18. Minor GC vs Major GC vs Full GC

Minor GC cleans the Young Generation

Major GC cleans the Old Generation

Full GC cleans the Young and Old Generation

19. AdProxy Latency Issues

20. JVM Options to know
java -XX:+PrintFlagsFinal -version | grep HeapSize

uintx InitialHeapSize := 268435456 {product}
uintx MaxHeapSize := 4294967296 {product}


java -Xms256m -Xmx2048m

-XX:MaxGCPauseMillis=200

-XX:+PrintGCDetails -XX:+PrintGCDateStamps

21. Java Memory Leak

Hashmap keys without proper equals and hashcode

Maps or lists which are growing forever

22. Troubleshooting

UI Options: jconsole or jvisualvm.

Command Line:
jstat -gc -t processID 1s
jmap -heap processID
jmap -dump:live,format=b,file=heap.bin processID
jhat heap.bin

JVM Options: -XX:+PrintGCDetails-XX:+PrintGCDateStamps

23. Thank you!