MARLINDB - A HIGH-SPEED, SCALABLE, DISTRIBUTED DATABASE

摘要

With the explosive growth of the Internet, web services that offer real-time experiences to their users are faced with difficult problems, because handling an ever increasing number of clients leads to scalability issues. Handling those problems by using relational database management systems together with replication and caching improves the situation only by a small factor, that further decreases when the majority of operations are writing data. We propose a database system that can scale with ease and that can handle more users by adding additional nodes to the system. Our database performs almost as good when using write operations as when reads are predominant. Furthermore, to increase the overall performance of the system, the database automatically prioritizes access to the data according to the actual user query patterns.