Research On MongoDB Load Balancing Mechanism Based On Cold And Hot Data

With the rise of the concept of big data, NoSQL overcomes the shortcomings of traditional database, which provides an effective solution to the massive data processing. In many NoSQL databases, MongoDB not only has the advantages of non relational database, but also has the traditional database index function and query mechanism. Mongodb as a distributed document database, the data stored in the form of BSON/JSON document. This makes the whole data can adopt a unified data model, the flexible storage mode and a variety of document structure enable the extensive application and attention in government, education, medical and other fields.Take mongodb as the research object, by reading part of the source code, analysis the storage architecture, focuses on the mongodb Auto-Sharding mechanism, and for the block splitting strategy and load balancing strategy in-depth study. Mainly completed the following work:1. Auto-Sharding mechanism of MongoDB in the block splitting strategy is based on the amount of the document data blocks the average split, which did not consider the data access time, which caused the difference between the server access load. Aiming at this problem, a block splitting strategy based on cold and hot data is proposed. The strategy is based on Naive Bayesian algorithm to classify the data, and introduces the definition of thermal load. The improved strategy is verified by experiments.2. Because the mongodb slice based on chunk split, the existing load balancing strategy by chunk of data migration to better ensure sub clusters in each slice server to store and maintain the amount of data to achieve a relatively balance, but there are ascending the tabs are caused by the load imbalance problem. In view of this problem, the MongoDB load balancing strategy based on hot and cold data is proposed, and the data migration time, migration object and migration target are solved according to the new Auto-Sharding mechanism. The effectiveness of the new strategy is verified by experiments.The research work shows that the mongodb Auto-Sharding mechanism block splitting strategy and load balancing strategy research and improvement of can effectively reduce the mongodb data access delay, and improve mongodb clusters of data throughput and so as to improve the overall performance of the system.