Atomic Classes

Introduction

As part of your senior engineer interview, you may be quizzed on the atomic classes and their utility. Though this is an advanced topic, understanding of it broadens your depth about, and command over concurrent systems and their workings.

So far we have seen locks that allow shared data to be manipulated safely among multiple threads. However, locking doesn’t come for free and takes a toll on performance especially when the shared data/state is being contented for access by multiple threads.

Cons of Locking

Locking comes with its downsides some of which include:

Thread scheduling vs useful work

JVM is very efficient when it comes to acquiring and releasing a lock that is requested by a single thread. However, when multiple threads attempt to acquire the same lock, only one wins and the rest must be suspended. The suspension and resumption of threads is costly and introduces significant overhead and this can be an issue for scenarios where several threads contend for the same lock but execute very little functionality. In such cases, the time spent in scheduling overhead overwhelms the useful work done. This is true of synchronized collections where the majority of methods perform very few operations.

Priority inversion

A higher priority thread can be blocked by a lower priority thread that holds the lock and itself is blocked because of a page fault, scheduling delay etc. This situation effectively downgrades the priority of the higher-priority thread to that of the lower-priority thread since the former can’t make progress until the latter releases the lock. In general, all threads that require a particular lock can’t make progress until the thread holding the lock releases it.

Liveness

...