hashed-wheel-timer | High Performance Timer / Scheduler Library compatible

hashed-wheel-timer is a Java library. hashed-wheel-timer has no bugs, it has no vulnerabilities, it has build file available, it has a Weak Copyleft License and it has low support. You can download it from GitHub, Maven.

Hashed and Hierarchical Wheels were used as a base for Kernels and Network stacks, and were described by the freebsd, linux people, researchers and in many other searches. Many modern Java frameworks have their own implementations of Timing Wheels, for example, Netty, Agrona, Reactor, Kafka, Seastar and many others. Of course, every implementation is adapted for the needs of the particular framework. The concept on the Timer Wheel is rather simple to understand: in order to keep track of events on given resolution, an array of linked lists (alternatively - sets or even arrays, YMMV) is preallocated. When event is scheduled, it's address is found by dividing deadline time t by resolution and wheel size. The registration is then assigned with rounds (how many times we should go around the wheel in order for the time period to be elapsed). For each scheduled resolution, a bucket is created. There are wheel size buckets, each one of which is holding Registrations. Timer is going through each bucket from the first until the next one, and decrements rounds for each registration. As soon as registration's rounds is reaching 0, the timeout is triggered. After that it is either rescheduled (with same offset and amount of rounds as initially) or removed from timer. Hashed Wheel is often called approximated timer, since it acts on the certain resolution, which allows it's optimisations. All the tasks scheduled for the timer period lower than the resolution or "between" resolution steps will be rounded to the "ceiling" (for example, given resolution 10 milliseconds, all the tasks for 5,6,7 etc milliseconds will first fire after 10, and 15, 16, 17 will first trigger after 20). If you're a visual person, it might be useful for you to check out these slides, which help to understand the concept underlying the Hashed Wheel Timer better. The early variant of this implementation was contributed to Project Reactor back in 2014, and now is extracted and adopted to be used as a standalone library with benchmarks, debounce, throttle implementations, ScheduledExecutorService impl and other bells and whistles. For buckets, ConcurrentHashSet is used (this, however, does not have any influence on the cancellation performance, it is still O(1) as cancellation is handled during bucket iteration). Switching to the array didn't bring change performance / throughput at all (however, reduced the memory footprint). Array implementation is however harder to get right, as one would have to allow multiple strategies for growth and shrinking of the underlying array. Advancement would be to implement a hierarchical wheels, which would be quite simple on top of this library.