This is an article on placing a task into a grid (cluster) where it is
picked up and processed by a worker, based on a deterministic
No task is lost; when a node crashes, the remaining nodes again
deterministically pick a worker which processes the task again.
There is no single point of failure, and nodes can be
added (e.g. to accommodate more load) and removed without interruption of the system.
The article can be found at
and here (HTML).
The source code is here.