Αρχιτεκτονική και αλγόριθμοι δρομολόγησης στο υπολογιστικό πλέγμα

The grid computing appears as a new model for the common use and aggregation of geographically scattered resources, such as personal computers, stations of work, sectors of computers and scientific instruments, aiming at the resolution of complicated and big scale problems in sectors such as science, industry and applied mechanics. Nevertheless, the growth of applications, the management of resources and the routing of work continue to constitute a challenge in the world community of the Grid growth. This paper is constituted by six chapters and an annex. Each chapter studies and analyzes a subject relative to the Grid Computing. The first chapter introduces Grid Computing. It also explains some definitions, its characteristics, the forms and the types of Grid. The first chapter also determines who are the users of this technology as well as the breadth of possibilities that a GRID system can cover. The second chapter is an extensive report of the fundamental significances and subjects of computing grids, as are the nature, the role and the main characteristics of the nominal organisms that guide in the determination of a particular, suitable architectural materialization and management of the nominal organisms and resources between the participants. This architecture follows the logic of "stratification" and is completed in this work with a summary description of the package GT2, with the intention to show how this architecture is applied in practice. The third chapter studies a new architecture that adopts the model of "stratification" and is directed in the network services that propose the OGSA and compares the two infrastructures OGSI and WSRF. The fourth chapter deals exclusively with the Globus Toolkit 4, which is constituted by various sections of software that focus in the growth of services according to the model of the web services. This chapter analyzes the way with which GT4 can be used in order to acquire an application access in the nodes of the grid, to submit them workloads and to accept notifications of the situation of each implementation and all of this happen in a secure environment. The fifth chapter describes various algorithms of allocating work in the calculating grid (scheduling algorithms) with the intention to study the quality of service that can be achieved from the various ways with which the allocated works will be distributed in the grid’s nodes. The algorithms of routing the work are separated in two categories. The dynamic algorithms, that are also separated in two categories, the "direct" (immediate mode) and "at bunch" (batch mode) and the static. For the study of the performance of the mentioned before algorithms, simulations have been done in environments of various degrees of heterogeneity as for the technical characteristics of calculating machines that compose it (machine heterogeneity) and as for the calculating needs of work that will be required to be executed (task heterogeneity). In the sixth chapter there is a short report in the big Grid project in Europe, in what is reported to as first wave and in what is reported to as second wave. Finally the annex describes the way the packages of Globus Toolkit are installed in the client and server instruments and the suitable settings that should be done in the Globus Toolkit in order for it to function regularly