There has been quite a discussion around the development of concurrent applications for multi-core processors. This meme is interesting to me since I just wrapped up a project where I architected a system that processes high volumes of data (~10TB) using a Java-based distributed computing solution.

While many of the posts are related to the use of threads for concurrency, my project opted for using separate processes. This decision was made because of some JNI integration with C code, which could potentially take down the entire process and lose hours of potential processing should we encounter instability. Many of the discussions have focused on desktop applications and gaming, but there are also strong implications for corporate, inside-the-firewall solutions. It is good to be reminded how architecting for concurrency changes the way we need to think about our system.

Have a look at Erik Sink’s post that summarizes and links to the various discussions if you haven’t been following along.