Java concurrency
2 likes568 views
This document summarizes key features and updates related to concurrency in Java releases. It discusses basics of concurrency including Amdahl's law and challenges of concurrent programming. It then covers constructs in Java 5+ like atomic operations, locks, conditions, executors and collections. The document also summarizes the fork-join framework and parallelism features in Java 8 including streams API, common pool and completable future. It highlights updates to ConcurrentHashMap in Java 8 to support parallel operations.
Java concurrency
- 1. Java Concurrency Towards a better parallel balanced life … ● By ● - [email protected] - [email protected]
- 2. Java Releases JDK 1.5: - Generics -Concurrency -Auto-boxing J2SE 6: - Collection Framework enhancement -Drag and Drop -Improve IO support J2SE 7: - Concurrency Utilities - Swing, -Fork/Join Pool , LWC&HWC - G1 J2SE 8: - Lambda expressions - Stream Package -More concurrency Updates
- 3. Agenda ● Basics of concurrency ● Java Concurrent Package ● Reentanrant Lock / Performance ● Concurrent Collections ● Insight of Fork/Join ● Concurrency in Java 8 ● Concurrent Collections java 8 updates
- 4. Basics of concurrency ● Amdhal's law – Speedup <= 1/ (F + (1-F)/N) ● Age of multi-core processors and requirement of parallel programming in mainstream. ● Writing concurrent programing is hard and writing correct concurrent program is even harder. ● Research in the field has been lead by Java designers and the concurrency interest group
- 5. General Design Principles Its the mutable state stupid Design for immutability Make field final unless they need to be mutable Encapsulate to manage complexity Guard all variables in an invariant with same lock Document the synchronization policy
- 6. Java Concurrency Constructs ● In Java 5 a set of synchronisers was built giving more power than intrinsic locking ● Atomic operations using Compare and swap ● Non-Blocking Collections
- 7. Intrinsic locks vs Reentarant Locks ● Intrinsic locks are very low level constructs ● Reentarant Locks offers more features ● Fair queuing ● TryLock ● Explicit Condition Queue
- 8. Dead Locks avoidance with try locks Dead Lock can be avoided using the try locks ● Try locks can used for Probabilistic lock acquire of the lock and exponential back-off and spin lock ● Classic Transfer money Deadlock Problem Typical traffic problem.
- 9. Condition Queues ● Intrinsic locks can also work as condition queues with wait and notify ● Problem occurs when Intrinsic condition queue may be used with more than one condition predicate ● Reentarant locks provides explicit condition queues which offers more features ● Fairness in access of the lock which reduce contention ● Help us create queues per condition predicate
- 10. Condition Queues ● Where a Lock replaces the use of synchronized methods and statements, a Condition replaces the use of the Object monitor methods. ● Consider a case of bounded buffer – final Condition notFull = lock.newCondition(); – final Condition notEmpty = lock.newCondition(); ● ArrayBlockingQueue provides the functionality ● A typical consumer/producer problem with BlockingQueue.
- 11. Performance Can Reentarant lock replace synchronized? – No , There are big issues when used carelessly – Use only when Fair queuing, probabilistic locking and explicit condition queues are required
- 12. Java Synchronizers ● Executors – Execution tasks in thread pools. – Cyclic barrier ,Future task ,Semaphores ,Countdown latch ● Non-blocking algorithms – AtomicInteger – Compare and Swap (CAS) – Writing a correct non-blocking algorithm are tough. ● Find out the code.
- 13. Concurrent Collection ● Copy on write array List ● Concurrent hash-map ● Blocking queues and producer – consumer pattern
- 14. Fork/Join Framework ● New parallel mechanism for compute intensive tasks, the fork-join framework. ● Runs on work-stealing algorithm. ● Approach – Partition into subproblems. – Create subtaks. – Fork subtasks. – Join subtaks. – Compose solution.
- 16. Java 8 way
- 17. Parallelism Built In ● Parallelism is the feature built in Java 8 ● U can iterate,filter, sort ,map reduce an array in parallel ● The streams api makes heavy use of the Fork- Join framework ● The entire set of features offered by the streams api can be made to work in parallel by just asking for the parallel stream ● The inherit complexity of Fork-Join framework is handled under the hood
- 18. Java 8 Concurrency Changes ● Addition of fork-join common pool ● ConcurrentHashmapv8 which supports bulk operations in parallel ● Completable future ● Changes to the Future Task implementation
- 19. Concurrent Hashmap ● Started as a replacement to Hashtable. ● Implements Lock stripping ● Doesn’t throw concurrent modification exception ● Java 8 concurrenthashmap does a lot ● Built in support for parallelism ● More in the code .