Sela Open House: Design and Architecture of Concurrent Applications

Yesterday I delivered another Sela Open House in Haifa (kudos to Philips MS for hosting this session). The subject was “Design and Architecture of Concurrent Applications”, and indeed much to my surprise I managed to avoid firing up WinDbg or writing lots of code, and instead talk about high-level principles.

Some of the things we covered:

• Hardware trends—the increasing number of processors means new challenges such as eliminating false sharing, reducing synchronization to a minimum, coping with NUMA architectures
• Types of concurrent applications—partitioning, pipeline
• Trends in software concurrency frameworks—the move from threads to tasks
• Types of concurrent algorithms—task parallelism, data parallelism, algorithms that don’t yield easily to parallelization

You can download the slides and sample code from here (caters for C# developers, but there’s an abundance of C++ ConcRT samples on the web as well). I also said I will provide some links for the material we didn’t cover on concurrency design patterns and debugging, so here goes:

• Debugging resources (includes a bunch of links to my posts and other locations, deals with debugging multithreaded applications but not only that)
• Some concurrency design patterns: thoughts on TimedLock, immutability, spinlock, cyclic lock-free buffer, lock-free operations, safe/unsafe cache, thread-local cache, read-write cache