Why Paradise?

With only minimal modifications, completely independent programs can share data. Consequently, programs with different functions can be developed independently. When one program changes, it need not have any effect on others with which it happens to share data. Paradise brings modularity at the application level to ensemble computing.

Virtual shared memories can continue to exist even after their creating programs terminate.

Paradise includes a sophisticated fault-tolerance facility which allows programs to guarantee data integrity despite hardware failures, abnormal process termination, lost connections between computers, and other failures. It is structured around a begin-commit transaction strategy similar to that used by state-of-the-art database systems. Changes (atomic transactions in database parlance) to a VSM are not made permanent until they are committed by the application performing them, and uncommitted transactions are cancelled in the event of failures.

Programs running on distinct or even incompatible computer architectures can access a common VSM without any special handling. Paradise takes care of all necessary data conversion automatically.

Programs sharing data need not run concurently, and applications can detach from and reattach to VSMs as desired.

Paradise includes facilities for controlling access to sharable VSMs above and beyond those provided by the host computer systems. Every use of a VSM requires permissions that can be precisely controlled by the creator of the VSM or other authorized users.

By combining Paradise with one of SCAI's other parallel computing environments, installations can use a single approach for both parallel and distributed computing. And like Linda, Paradise enables data sharing and computing capacity allocation throughout an entire enterprise.

Paradise can be used to create so-called live libraries: servers which can perform various standard types of computations for any client which requests them. Unlike traditional subroutine libraries which have been parallelized for a single architecture, such servers can run in parallel on any available computer systems. This allows, for example, a sequential application on a mainframe to take advantage of a specialized multiprocessor (or of the idle CPU capacity in a local area network) to perform computationally-intensive operations. This is a another way in which available computing resources can be dynamically deployed on an enterprise-wide basis using SCAI products.

In the software business since 1980, SCAI has a well-earned reputation for solid commercial products, combined with excellent support, training, and consulting services.