Conception and research results of Distributed Computer Systems (DCS) with programmable structure are stated. Functioning models and continuous methods are described to calculate indices of DCS efficiency. Results of analysis and synthesis of DCS structures are given. Stochastic strategies are suggested to optimize DCS functioning in parallel multiprogramming modes. Functional structures of embodied DCS ("Minsk-222", MINIMAX, SUMMA, MICROS and geographical-distributed multicluster system) are considered. It is showed the given conception let us create technical-economic effective and robust supercomputers.

In work the mathematical model for calculation of indices of realizability of solving tasks of stream with refusals on the distributed computing systems (CS) is offered. Researching a transitive mode of functioning CS; analytical formulas for calculating indices of realizability of solving tasks are derived.

Modern tools for information processing are distributed computer systems (DCS) of high performance. They are MPP-systems and contain high number of processors (up to 1 000 000). The exploitation of such systems is always connected with financial expenditures. The project contains examination of a continuous method of calculation of technical and economic indices of functioning distributed CS (income and expenditure functions) in stationary and transient condition.

Interconnection networks of modern distributed computer systems are now hierarchical. In such systems communication time between processor cores depends on a placement of processors in a system. For example, in large-scale SMP computer clusters first two levels are formed by switches of two-stage fat tree and a third level is presented by a shared memory of computer nodes. In this paper, we describe a dynamic optimization method for collective communication operations on hierarchical computer clusters. Our approach exploits knowledge of the L-level hierarchy and is based on a mapping of intensively communicating processes into the same computer node. Performance results on multicore SMP computer clusters with In?niBand and Gigabit Ethernet networks indicate that the Allgather algorithms based on our approach outperform the original Allgather algorithms.

The algorithms of functioning optimization of distributed computer systems at scheduling moldable jobs with allowance for the penalty for delay of its decision and priority of possible choice of subsystems configurations are proposed. Algorithms modeling results are presented.

Parallel job streams servicing problem in geographically-distributed computer systems is considered. Functional structure of developed program tools of decentralized service of parallel MPI-jobs in multicluster computer systems are described. Results of experiments on multicluster computer system are represented.

An approaches to parallel programs incremental and differential checkpointing are stated. Hash based sequential and parallel algorithms are proposed and investigated. An efficiency of proposed approach for iterative and direct methods for solving complex tasks are shown. Good scalability of parallel version of proposed algorithm is shown.

Weaklyuniversal VB coding is proved to be for stationary sources set. Necessary and sufficient conditions of existing universal VB coding of stationary sources set owe obtained.

Architecture principles and software of geographically-distributed computer systems are described. A real configuration of geographically-distributed multicluster computer system is presented. This system has been created and developed by the Computer centre for parallel technologies of Siberian State University for Telecommunication technologies and Information techniques in conjunction with A.V. Rzhanov Institute of semiconductor physics of Siberian Branch of Russian Academy of Science. The description of the system software (including open source components and extensions developed by collective of the author's scientific school on distributed computer systems) is resulted.