Scaling Up with STM.NET (Software Transactional Memory)

Every developer that spent enough time with concurrent applications must have learned that producing a reliable, high performance application that scales gracefully to many core hardware is extremely difficult to do.

In order for the application to scale and maintain high throughput, we have to design the threading model such that at any given time the application will produce the optimal amount of threads that will put the maximum number of cores to work with minimum contentions. In order to keep the amount of contentions low, besides of not over introducing threads, it’s crucial that we’ll find an efficient way to protect the data that is being shared among the threads from concurrent access.

In order to make parallel programming easier, .NET framework 4.0 introduced the TPL (task parallel library) together with new parallelism constructs that facilitated the process of adding concurrency to applications. However, although more and more code is tempted to run in parallel, we get no real help from the runtime with protecting the shared data from concurrent access. Applications that require coarse-grained isolation can get away relatively easy by utilizing techniques such as message passing, which deters sharing and eliminate the need for locks almost entirely. However, in order to achieve greater scalability there’s not escape from using fine-grained locks, and so the problems begin…as we’ll find in the next chapter.

Read more: Design Codes

Posted via email from jasper22's posterous