The Repository of VTEI
id: 25605
Title: Conflict resolution in computer system
Authors: Yatskovska R.
Keywords: computer system, multicriteria problems, conflicts in computer systems, optimal strategy, deadlocks, prevention of deadlocks
Date of publication: 2020-12-16 15:38:02
Last changes: 2020-12-16 15:38:02
Year of publication: 2020
Summary: In the network during data transmission there are uncertain situations that prevent the receipt of data packets to the destination nodes. Therefore, the problem of resolving deadlocks is one of the most important in the design of data transmission (especially with buffering) in computer networks. There are two traditional approaches to solving deadlocks: networks and schemes are designed that eliminate deadlocks; development of a distributed algorithm that detects a deadlock situation and then derives the network this situation using some permitted distributed algorithm. The proposed approach is based on providing indivisible resources to processes in such a way as to minimize losses due to conflicts. The multicriteria problem of providing indivisible resources to processes is investigated, and the principle of optimality is expressed by the known binary relation on the set of average vectors of penalties for conflicts on each of resources. It is shown that the joint use of the apparatus of choice theory and the classical apparatus allows to expand the known problem statements by using more general principles of optimality; examples of solving multicriteria problems of optimal conflict resolution management in computer systems are given. Quantitative estimates of the gain at the chosen optimal strategy of conflict resolution in multiprocessor computer systems are obtained.
URI: http://sel.vtei.edu.ua/card.php?id=25605
Publication type: Стаття у закордонних наукових виданнях
Publication: Scientific Letters of Academic Society of Michal Baludansky. 2020. Vol. 8, № 4. С.147-149
In the collections: Статті/ Видання інших установ/
Published by: Адміністратор
File : 25605.pdf Size : 2011064 byte Format : Adobe PDF Access : For all

Enlarge