Real-time data management in data grid systems

scientific and commercial applications are increasing. In order to meet these requirements, Data Grid systems that orchestrates thousands of processors under a common application framework are developed. Data intensive applications that require large datasets to be processed, comes with real-time requirements withincreasing frequency. These applications are called real-time applications and they come with a predefined deadline before which the applications should be completed. The performances of Data Grid systems that runs real-time applications are affected by the underlying job scheduling, data dissemination, data replication and advance reservation mechanisms. A Data Grid system model that incorporates all these components is presented in this study. In the proposed model, services that constitutes the Data Grid system and inter service interationsare defined. In the proposed model, hierarchical job scheduling, hierarchical data dissemination, pull based, push based distributed and push based centralized data replication models are supported. Furthermore, in the proposed model, an advance reservation system is devised to enable advance reservation of the system resources. Three data access profiles (random, geometric, and zipf) and two data organization models (federative and hierarchical) are defined. Rand, EDF, MCTF,MCwDP, MMwDP job scheduling algorithms are proposed and performances are measured. The problems of satisfying real-time data transfer requests via one (RTU/DDP) or more paths (RTS/DDP) are introduced. SP_MinHop, SP_MinDelay, SP_MinMin, SP_MinCon, MinHop/FPF, MinDelay/FPF, MinMin/FPF, MinCon/FPF, IO_UFF, MNOFF, MOFF, GA, kSP, kDP, ESMP, BSMP heuristic algorithms are proposed for data dissemination problem andperformances are measured. Algorithms for various data replication models are developed and performance measurements of the proposed MRD, RT_DIW and RT_CENT algorithms are shown.