How is SingleThreadModel implemented in Tomcat?

Avi Kak

The question mixes together two rather independent aspects of a servlet container: "concurrency control" and "thread pooling".

Concurrency control, such as achieved by having a servlet implement the SingleThreadModel interface, addresses the issue of thread safety. A servlet will be thread-safe or thread-unsafe regardless of whether the servlet container used a thread pool. Thread pooling merely eliminates the overhead associated with the creation and destruction of threads as a servlet container tries to respond to multiple requests received simultaneously. It is for this reason that the specification document for Servlet 2.2 API is silent on the subject of thread pooling -- as it is merely an implementation detail. However, the document does indeed address the issue of thread safety and how and when to use SingleThreadModel servlets.

Section 3.3.3.1 of the Servlet 2.2 API Specification document says that if a servlet implements the SingleThreadModel it is guaranteed "that only one request thread at time will be allowed in the service method." It says further that "a servlet container may satisfy this guarantee by serializing requests on a servlet or by maintaining a pool of servlet instances."

Obviously, for superior performance you'd want the servlet container to create multiple instances of a SingleThreadModel type servlet should there be many requests received in quick succession. Whether or not a servlet container does that depends completely on the implementation. My experiments show that Tomcat 3.1 does indeed create multiple instances of a SingleThreadModel servlet, but only for the first batch of requests received concurrently. For subsequent batches of concurrent requests, it seems to use only one of those instances.
0 Comments  (click to add your comment)
Comment and Contribute

 

 

 

 

 


(Maximum characters: 1200). You have 1200 characters left.

 

 

About | Sitemap | Contact