Abstract:
Today a lot of effort is worldwide put into development exaplops computers, which will be able to perform $10^{18}$ floating-point operations per second. Effective performance of the cluster on real tasks is considerably less of that value due to synchronisation costs of concurrent processes. This paper introduces a concept of passive cache, which eliminates any need for process synchronization in some tasks, and proposes a system architecture of high performance distributed DBMS, based on the passive cache. (In Russian).
Key words and phrases:parallel and distributed computing, very large databases.