Why you should use NIO for scalability

The key to scalability is NON-BLOCKING IO. Most operating systems support non-blocking I/O, which actually means that when you utilize an I/O resource for reading or writing (say the streams from a socket) there is no blocking operation. So if you read from a stream your read function would immediately return regardless of whether there is data available or not. In Java, non-blocking I/O is provided by the Java New I/O (NIO) library using Selectors and perhaps the Reactor pattern [F]  [F] A nice overview of NIO is athttp://gee.cs.oswego.edu/dl/cpjslides/nio.pdf. This has a major impact on scalability because the threads are only held as long as there is work to do. Once they’re finished with the available data, they are returned to the thread pool so that they may be reused for processing other requests. In this model the threads are allocated to connections only when data is available for processing, which inherently leads to better resource utilization.

Note: This is somewhat off-topic, but if you’re looking to do a lot of work with NIO and networking, we recommend looking at the Apache MINA project at http://mina.apache.org/. MINA provides some nice abstractions for NIO that allows you use a stateful approach to developing NIO applications without having to deal with a lot of the underlying details of using NIO.