Институт системного программирования им. Иванникова РАН 2 Московский физико-технический институт Аннотация. Рассматривается компонентный подход к созданию систем и сайтов из готовых ресурсов (компонентов, объектов, сервисов и reuses). В основе подхода лежит графовая и компонентная модель (КМ), включающая функциональные, системные, сервисные и интерфейсные программные ресурсы и алгебру их обработки. Функции объектов-компонентов описываются в языках программирования (ЯП), а их интерфейсы в языках API, IDL, WSDL и др. Они реализуются в клиент-серверной архитектуре и взаимодействуют между собой через интерфейс. Дан анализ моделей систем, веб-сайтов и приложений, выполняемых в двухуровневой клиент-серверной архитектуре и представлен современный путь развития этой архитектуры в связи с увеличением числа пользователей, работающих с современными огромными объемами данных Big Data и Cloud Computing. Готовые ресурсы проверяются на правильность их спецификации в ЯП, тестируются и собираются сведения об ошибках для их исправления и оценки надежности. Отработанные ресурсы конфигурируются и система проверяются на функциональность. Описан сайт http://7dragons.ru, включающий готовые ресурсы и артефакты технологии разработки программных систем и новые сервисно-компонентные ресурсы (SOA, SCA, SOAP и др.), используемые для моделирования веб-систем и сайтов.

The work is dedicated to the topic of various (both static and dynamic) methods of balancing the computational load with parallel implementation of algorithms on shared memory. Attention is also paid to the topic of parallelizing programs in especially difficult cases-when the used algorithm is purely sequential,

We consider a distributed system based on a strongly connected directed numbered rooted graph without multiple edges and loops in which the arcs outgoing from the vertex are numbered from 1 to the out-degree of this vertex. The task is to ensure the message passing from each sender to each receiver by the shortest paths. For this it is proposed to construct at each vertex a of the graph a mapping that assigns to each vertex number b the arc number with which the shortest path starts from a to b. This mapping is useful for solving many problems, in particular, the task of self-configuration a program-configurable network. The network switch stores in program-configurable tables the mapping of the address of the target host to the number of the output port. The time for passing packets is minimal, if the shortest paths are used in the mapping. Another problem is the simulation of distributed algorithms developed for distributed system based on an undirected connection graph on a class of systems based on a directed connection graph. Here we have to solve the backtracking problem: instead of passing a message from the end of the arc a→b to its beginning, the message is passing along the directed path b→ ... →a. It is assumed that the mapping construction algorithm should start and end at the root of the graph. Two algorithms for constructing a mapping are proposed: a "fast" algorithm with estimates T = O(n) and N = O(n) and an "economical" algorithm with estimates T = O(n 2) and N = O(1), where T is the algorithm running time, N is the number of messages simultaneously transmitted along the arc, n is the number of vertices of the graph. It is proved that these estimates are lower bounds on the class of algorithms that correctly operate on the class of all graphs under consideration.

The paper is devoted to deriving tests with guaranteed fault coverage for Finite State Machines (FSM). FSMs and their compositions are widely used when deriving tests for various systems, including interactive services, and the synchronous composition describes the behavior of a multi modular system when all the channels are active at each time instance. Without loss of generality, we assume that an output port of each FSM component is connected with at most one input port of another FSM component and propose a simplified approach for deriving the synchronous composition of deterministic complete FSMs using well known operators over FSMs. We also extend the sufficient conditions for checking the closure of the synchronous composition operator with respect to deterministic complete FSMs showing that for the operator closure it is sufficient that each composition cycle contains a Moore pair, i.e., a pair of some component FSM such that outputs do not significantly depend on inputs (Moore condition). If a system under test has the same structure as the reference composition and faults of component FSMs do not break the Moore condition, then a transition tour of the composed FSM is a good quality test when component states can be observed (testing with open state). We also propose how to derive a test with guaranteed fault coverage for a component FSM of our interest when states cannot be observed under the condition that input and output ports of a component under test can still be observed.