Hypercomputer

Hypercomputer is a network-based parallel computer system. The idea is to connect the processors of the computers into a local area network (LAN). It is an option for companies that require the computing power of supercomputers but only occasionally.

The given problem is broken into small pieces and each piece is given to each of many processors. This networking arrangement intends to make use of the idle processors effectively.

Linda, for example, is a software available to use a hypercomputer. It is a set of parallel programs developed by David Galernter, Nicholas Carriero, Jerrold Leichter and others from Yale University. Linda makes use of any processor that is available with the help of tuple space. Tuple space is a shared memory where the items have no address so that any task can be accessed irrespective of its physical location of the items. It stores the input data and intermediate results from completed operations.

Hypercomputer uses a process known as daemon to distribute the tasks to be performed by the processors within its network. The daemon process runs continuously in the background. It has two parts: scheduler and allocator. The scheduler waits for processors in the network to become available. The allocator sends message to the scheduler if a processor is available for use.