The Solution Of Distributed Shared Memory Based On Domain Oriented Search Engine

Distributed Shared Memory technology, which is one of the main directions of computer parallel technology nowadays, has very expansive application foreground in server cluster, artificial intelligent and search engine. Especially in the direction of search engine, there are plenty of new breakthroughs in both theory and technology in recent years. The project called Domain Oriented Chinese AI Search Engine is researched recently, whose kernel algorithm needs to compute the high dimension metrices, which are not able to be read and written in solo server for the reason of limited memory. For solving the problem, combining the technical features of distributed shared memory technology and the practice of intelligent Chinese search engine, this paper proposes the Solution of Distributed Shared Memory Based on Domain Oriented Chinese AI Search Engine called nDSM(nutch Distributed Shared Memory System), which provides high capability and high efficiency virtual memory space and safe storage service.This paper emphasizes the design thinking and system framework of nDSM system. Based on the business requirement, the design idea is to realize a distributed shared memory system of easy to operate and extend. Specifically the system manages the operation of multithreading read and written on page relative matrices, and establishes on the united shared virtual memory, which is achieved both by interface server and subserver group, when the cache consistency, memory consistency and server load balance strategy are reached. According to the design thinking, the nDSM system is transparent to other parts of the search engine, and shows certain loosely coupled from other aspects of the main system for the reason of autonomy function of nDSM system.By analyzing the functions of the Domain Oriented Chinese AI Search Engine, this paper concludes that this system has a theoretical stable efficiency, and could satisfy the requirement from the kernel algorithms for high dimension matrices.