alecmev / task-sprayer

Allows to distribute and run separate quanta of tasks on separate nodes of a cluster. Technological background - Java, Linux; server runs on a local master node and listens for potentially unlimited amount of clients; when a client is connected successfully, a queue of tasks is polled and a data set is transferred to the client; when the client is finished calculating the result, it's downloaded back and stored in a file; in case of a failure, the task is submitted to another node; after N fails the task is taken out of the main queue and is stored in a special place. First version was implemented in pure C (folder #backup).