Programming WCF Services, 2nd Edition

Streaming

By default, when the client and the service exchange messages, these messages are buffered on the receiving end and delivered only once the entire message has been received. This is true whether it is the client sending a message to the service or the service returning a message to the client. As a result, when the client calls the service, the service is invoked only after the client's message has been received in its entirety; likewise, the client is unblocked only once the returned message with the results of the invocation has been received in its entirety.

For sufficiently small messages, this exchange pattern provides for a simple programming model because the latency caused by receiving the message is usually negligible compared with the message processing itself. However, when it comes to much larger messages—such as ones involving multimedia content, large files, or batches of data—blocking until the entire message has been received may be impractical. To handle such cases, WCF enables the receiving side (be it the client or the service) to start processing the data in the message while the message is still being received by the channel. This type of processing is known as streaming transfer mode. With large payloads, streaming provides improved throughput and responsiveness because neither the receiving nor the sending side is blocked while the message is being sent or received.

I/O Streams

For message streaming, WCF requires the use of the .NET Stream class. In ...