Transaction Planning Methods in Hyperconverged Architecture Systems

Abstract

The analysis of the features of the functioning of systems with hyperconverged architecture is carried out. Transaction efficiency in such systems is reduced compared to decentralized systems. The purpose of the research: to develop a method for planning transactions in systems with a hypeconvergent architecture, which will take into account the specifics of the functioning of such systems. The development of the method takes into account the centralized management of the transaction package and the distribution of various resources. Existing methods for determining the sequence of transactions in systems with hyperconverged architecture are considered. The methods that were considered were based on the greedy, clustering, and ant algorithms. For each method, its features and functioning scheme are determined. Analysis of existing methods showed the advantages of the greedy algorithm with a small system load. It is also proved that with the growth of information volumes and the number of simultaneously executed transactions, each of the methods considered is less effective than with decentralized management. Therefore, a method for planning the execution of transactions through the sharing of these optimization algorithms is proposed. This allowed to reduce the execution time of the optimal transaction plan in comparison with existing methods. The experiments showed that the effectiveness of the proposed method increases with increasing amount of information. Which is processed by a transaction package of a hyperconverged system.