R.Sh. Minyazev

Comparing different approaches for distribute queries between clusters-components in multicluster DBMS of conservative type. Searching the best, which provide close to the minimum mean-statistical response time to new incoming query. Comparison bring through position of estimation of statistical parameters M and σ for time series, which formed during processing of sequences of queries with limited length M using different types of routing. To analyze case of multicluster DBMS continuous work, when M - unlimited, modeling an abstract system work. It handled three sequences of queries, each have length M=1000, they corresponding to the three laws of distributions: uniform, normal and poisson. Different approaches to distribute between monoclusters used. To assess the distribution of delays in case of varying the number of users N and count of monoclusters n comparing coefficient of variation V=σ/M for received time series. The study concluded preference of circular routing regardless of the law of distribution of queries in sequences. In the presence of combination in processing different queries parameter k=2. For systems without combination k=1. Detected positivism of using external queue in module ROUTER