Aggregation arrangement features of network interface channels in multiprocessor computing systems

Abstract

Listen

Translate article

The paper identifies ways to increase the multiprocessor computing system performance by reorganizing the architecture of its network interface. It is shown that the computational parallelization performance significantly depends on many factors, the most essential of which is the data transfer between the boundary nodes of a multiprocessor system, which is the algorithm's slowest part and can significantly reduce the effect of increasing the number of processors. Hence, it was established that improving the multiprocessor systems performance by reorganizing the network interface structure is relevant interesting nowadays, and its study is at the active development stage. The research aims at the aggregation arrangement of network interface channels in multiprocessor computing systems. It is shown that the main channel aggregation mode advantage is that the data interchange speed is significantly increased, as well as the reliability of the cluster system.