APPROACHES ТО EFFECTIVE IMPLEMENTATION OF ITERATIVE ALGORITHMS ON MAPREDUCE MODEL
Iterative algorithms are an important class of tasks, and are found in many areas, such as data-mining, machine learning, reference analysis and other. The most well-known and often applicable tool in big-data computing is the MapReduce model and in particular its open source implementation Hadoop. This paper provides an overview of existing approaches to the implementation of the MapReduce-like models supporting the effective implementation of iterative algorithms. It is shown that the main factors affect adversely the performance of iterative algorithms on MapReduce model is the disk and network IO, which in most cases can be prevented by various modifications of the model and software environments that implement it.
Authors: A. V. Mityakov, Y. S. Tatarinov
Direction: Informatics, management and Computer Technology
Keywords: Big-data, data computing, MapReduce, iterative algorithms
View full article