Selection of the optimal number of tables for parallel hashing with limited hardware resources of distributed computing systems

In distributed computing systems, the actual problem is to organize data exchange between computing nodes in a network, and the network used bridges or switches. The main function of the bridge is to filter local traffic. Local traffic, ie frames to the network connected to the first port, should not be passed over the bridge, that is to be filtered. To make a decision on the need to filter the frame bridge organized self-learning filter table. Algorithms for filtering and filling tables decision algorithm to filter frame specified standard IEEE 802.1D. The main results in the field of filtration belong to large corporations such as XEROX, IBM, Cisco, Alcatel, AMD, Intel, and others.

This article discusses features of the parallel hash tables for the organization of personnel filtering in distributed computing systems. The proposed method of organizing the tables filtering to reduce the processing time frames in the internetwork bridges and switches to ensure a low probability of overflow table filtering. the proposed method of organizing the tables in the firewall filter frame bridges and switches can significantly reduce the amount of memory required to store the table filter with comparable probability of overflow. The performed statistical studies have identified the optimal number of parallel tables for a given amount of memory allocated to store them.