Scalability Evaluation of the GIPSY Runtime System

Intensional programming is a declarative programming paradigm that is suitable for scientific programming since it allows natural expression of equations regarding multidimensional objects or concepts evolving in a multidimensional context so that the simplicity of these equations are kept. The General Intensional Programming System (GIPSY) project aims at providing a software platform for the long-term investigation of intensional programming. The GIPSY consists of a flexible compiler and a scalable runtime system, where the compiler translates any flavor of intensional program into source-language independent runtime resources, and the runtime system uses the runtime resources to execute the program in a demand-driven and distributed manner, i.e. computation requirements are wrapped into demands and are distributed among networked computers, so that the computations can be executed distributively and concurrently to shorten their overall computation time. The multi-tier architecture adopted for the GIPSY runtime system is for research goals such as scalability. It consists of the Demand Generator Tier that generates demands, the Demand Store Tier that stores and dispatches demands, as well as the Demand Worker Tier that computes demands. All the tiers are allocated in registered computers called the GIPSY nodes, and all the GIPSY nodes and tiers are under the management of the General Manager Tier, with which new nodes can be registered and new tiers can be allocated at runtime to deal with increasing workload. This thesis covers the development of the scalable GIPSY runtime system using the multi-tier architecture, and presents the assessment of the scalability of the developed GIPSY runtime system