| With the rising of cloud computing and big data, especially in recent years,massive scale application level data continue to emerge, the rapid growth ofunstructured data storage and processing needs, these new massive data with thenew demand for data storage. While traditional relational database has not been toadapt the Web2.0sites, especially the needs of ultra large scale and high concurrencySNS type of site, the cost of maintaining the relational databases is increasing on themajor Internet companies and network storage company. Because relational databasedesign issues such as a single data structure, relatively complex query process,exposed some insurmountable problems, the urgent need of a new high-performance database storage to supplement or replace the relational database. TheNoSQL database due to their advantages, such as high performance, highavailability, scalability, and flexible data model, has been developing rapidly, theperformance analysis and tuning unstructured distributed storage system has becomea hot research technology.Firstly, on the basis of searching for relevant theories of NoSQL technology,from the advantages of the characteristics of NoSQL databases, analyze storagesystem design and implementation guidelines to follow.By studying the currentmainstream NoSQL database storage system, the reference implementation offoreign NoSQL products, designed and implemented based on NoSQL storagesystems, the reference implementation of foreign NoSQL products, designed andimplemented based on NoSQL storage systems, and the using of system testingtools for the effective performance analysis and evaluation. The main researchcontents are as follows:(1)Through the NoSQL data model in the depth analysis and research to achievea data model based on key storage system that can support the basic commandqueries, support the completion of a series of commands with the basic functions ofa transaction mechanism, the use of design patterns observer ideology subscriptioninformation, support key for data storage used in multiple server clusters and otherfunctions, and the system made a detailed analysis of API functions.(2)In the depth analysis of the traditional NoSQL storage systems commonlyused data structures and algorithms problems through the improvement andperfection of some inefficient data structures and algorithms, to improve storage efficiency and reduce storage system CPU consumption, focusing on research andimprove consistent hashing by adding the virtual node to solve the balance problemsof traditional NoSQL storage system about consistent hashing algorithm.(3)Through the storage system performance test analysis and design to develop abenchmark tool menkv-benchmark, to solve the lack of test equipment and testequipment performance difference on NoSQL storage systems, resulting in a series oftest results through specific test data comparison average performance proves thesystem can handle15,000commands per second, higher than GlobalsDB performancestorage systems, proving the reliability and effectiveness of the system performance. |
