The object pool pattern uses a set of initialized objects kept ready to use, rather than allocating and destroying them on demand. A client of the pool will request an object from the pool and perform operations on the returned object. When the client has finished, it returns the object, which is a specific type of factory object, to the pool rather than destroying it.
Object pooling can offer a significant performance boost in situations where the cost of initializing a class instance is high, the rate of instantiation of a class is high, and the number of instances in use at any one time is low. The pooled object is obtained in predictable time when creation of the new objects (especially over network) may take variable time.
However these benefits are mostly true for objects that are expensive with respect to time, such as database connections, socket connections, threads and large graphic objects like fonts or bitmaps. In certain situations, simple object pooling (that hold no external resources, but only occupy memory) may not be efficient and could decrease performance.
Handling of empty pools
Object pools employ one of three strategies to handle a request when there are no spare objects in the pool.
- Fail to provide an object (and return an error to the client).
- Allocate a new object, thus increasing the size of the pool. Pools that do this usually allow you to set the high water mark (the maximum number of objects ever used).
- In a multithreaded environment, a pool may block the client until another thread returns an object to the pool.
When writing an object pool, the programmer has to be careful to make sure the state of the objects returned to the pool is reset back to a sensible state for the next use of the object. If this is not observed, the object will often be in some state that was unexpected by the client program and may cause the client program to fail. The pool is responsible for resetting the objects, not the clients. Object pools full of objects with dangerously stale state are sometimes called object cesspools and regarded as an anti-pattern.
The presence of stale state is not always an issue; it becomes dangerous when the presence of stale state causes the object to behave differently. For example, an object that represents authentication details may break if the "successfully authenticated" flag is not reset before it is passed out, since it will indicate that a user is correctly authenticated (possibly as someone else) when they haven't yet attempted to authenticate. However, it will work just fine if you fail to reset some value only used for debugging, such as the identity of the last authentication server used.
Inadequate resetting of objects may also cause an information leak. If an object contains confidential data (e.g. a user's credit card numbers) that isn't cleared before the object is passed to a new client, a malicious or buggy client may disclose the data to an unauthorized party.
If the pool is used by multiple threads, it may need the means to prevent parallel threads from grabbing and trying to reuse the same object in parallel. This is not necessary if the pooled objects are immutable or otherwise thread-safe.
Some publications do not recommend using object pooling with certain languages, such as Java, especially for objects that only use memory and hold no external resources. Opponents usually say that object allocation is relatively fast in modern languages with garbage collectors; while the operator
new needs only ten instructions, the classic
delete pair found in pooling designs requires hundreds of them as it does more complex work. Also, most garbage collectors scan "live" object references, and not the memory that these objects use for their content. This means that any number of "dead" objects without references can be discarded with little cost. In contrast, keeping a large number of "live" but unused objects increases the duration of garbage collection. In some cases, programs that use garbage collection instead of directly managing memory may run faster.
In the .NET Base Class Library there are a few objects that implement this pattern.
System.Threading.ThreadPool is configured to have a predefined number of threads to allocate. When the threads are returned, they are available for another computation. Thus, one can use threads without paying the cost of creation and disposal of threads.
Java supports thread pooling via
java.util.concurrent.ExecutorService and other related classes. The executor service has a certain number of "basic" threads that are never discarded. If all threads are busy, the service allocates the allowed number of extra threads that are later discarded if not used for the certain expiration time. If no more threads are allowed, the tasks can be placed in the queue. Finally, if this queue may get too long, it can be configured to suspend the requesting thread.