Рақамли технологияларнинг назарий ва амалий масалалари 9-том 3-сан (2026) · 99-111-беттер
Principles of organizing parallel algorithms in cloud computing environments
Исроилов, Ш.Й., Каршиев, З.А., Саттаров, М.А., Абдуваитов, А.А., Бурибоев, А.Ш.
Аннотация
The growing volume of big data in modern information technologies necessitates the development of efficient parallel algorithms in cloud computing environments. Existing systems such as Hadoop and Spark face challenges related to static task distribution and high memory consumption, respectively. This study proposes a hybrid parallel computing approach that integrates DAG (Directed Acyclic Graph)-based task graphs, MapReduce, and Fork-Join paradigms. The proposed CloudParallelCore v1.0 software platform features a five-layer architecture enabling cloud provider-independent operation. The hybrid scheduler algorithm dynamically distributes tasks through the priority function f(p, w) = α·priority + β·wait_time. Experimental results demonstrate that 69 out of 70 tests (98.6%) passed successfully, with test coverage exceeding 90%. In MapReduce mode, a speedup of 1.09× was achieved for a dataset of 10,000 elements using 4 worker processes. In Fork-Join mode, sorting 50,000 elements completed in 1.20 seconds. The single identified issue – a reduce phase timing measurement boundary for small datasets – had no impact on functionality. The results confirm the effectiveness of the hybrid approach for medium and large-scale data processing.
облачные вычисленияпараллельные алгоритмыMapReduceFork-JoinDAGгибридный планировщикCloudParallelCoreпланирование задачcloud computingparallel algorithms
Метадайындар булагы: журналдын OAI-PMH архиви · Sindex толук текстти сактабайт, булакка шилтеме берет.