An object pool implementation in Java which tried to overcome the shortcomings of existing pools:
- ability to control how objects are reused (last used, first used)
- provide statistics about pool operations
- provide information about pooled objects
The pool aims to be fast but at the same time provide a clean and flexible API. It borrows ideas, terminology and concepts from existing object pools but at the same it tries to provide what's missing from existing pool implementations.
One of the major shortcoming of existing pools (in the opinion of the authors) is the strategy around optimal number of pooled objects. Most pools implement a minimum idle objects concept, which is supposed to help in terms of performance since there will be always a number of objects available to be borrowed but also never keep more idle objects that minimum idle objects. However, this strategy creates in many cases an over provisioning situation when the total number of objects across a given number of services is too high and services will not be able to create new objects.
The main goal of the pool is to provide objects as fast as possible but without creating more than a maximum number of objects. Desired hints should be provided to the pool but the most efficient way to provide these objects without wasting resources should be provided.