Sibiryakov M.A., Vasyaeva E.S. —
Modification and Modeling of Data Processing Algorithms in Cache Memory of Data Storage Systems
// Cybernetics and programming. – 2016. – ¹ 4.
– P. 44 - 57.
DOI: 10.7256/2306-4196.2016.4.18058
URL: https://en.e-notabene.ru/kp/article_18058.html
Read the article
Abstract: The present article is devoted to the question about increasing productivity of the cache memory subsystem of data storage systems. The main purpose of the article is to increase the speed of executing the basic algorithms of the information search operation in controlling index structures. The subject of the research is the controlling index tables the execution of the basic algorithms is based on. In their article the authors offer to execute modified data processing algorithms and index based on a unique method of hashing. The authors provide results of the analytical modelling of initial and modified data processing algorithms using the method of Markov chains. The authors evaluate the average performance of these algorithms. They also carry out a computer-aided simulation modelling of the search operation within the data structures under research. Within the framework of the studied method of controlling the cache memory, the authors prove that it is reasonable to use hash tables in order to build controlling index tables that containt a great number of stored messages. The research shows that implementation of hash tables allows to significantly increase the speed of the basic data processing algorithms in the cache memory of data storage subsystems.
Sibiryakov M.A., Sukhikh A.V., Ivanov K.V., Koshpaev A.A. —
Building a computing cluster based on the PCI Express communications environment
// Cybernetics and programming. – 2015. – ¹ 5.
– P. 173 - 180.
DOI: 10.7256/2306-4196.2015.5.17035
URL: https://en.e-notabene.ru/kp/article_17035.html
Read the article
Abstract: The article discusses the implementation of a productive and reliable cluster with low cost. As a solution, the authors recommend building a cluster system based on the PCI Express data input / output protocol. The main goal is to find ways to increase the productivity of the cluster system and its circuit protection. The subjects of the study are the algorithms for data transfer and processing on the switch, algorithms for search and displacement data in storage system, switch ports arbitration scheme. The research methodology includes structured systems analysis in the domain of the connection structures, the construction of cache storage, methods of constructing cluster systems. The authors propose the following structural and algorithmic solutions: a modified algorithm for address translation of the external PCI Express Switch; an improved method and algorithm for data searching and processing in the cache data storage systems based on a unique hash; formalization of all stages of arbitration algorithms for ports and virtual switch channels.