The improvement of auto-scaling mechanism for distributed database - A case study for MongoDB

In recent years, cloud computing is the most popular topic on the IT industry. The underlying virtualization technologies, that make cloud computing possible, also get more and more attention. Gradually, companies move their services to the virtual host. These services include: desktop virtualization, application virtualization and database virtualization etc. Among these services, database virtualization can improve flexibility, maximize efficiency, lower costs and ease administrative overhead. In this paper, we use on-demand features of cloud computing and sharding features of MongoDB to provide an auto-scaling database virtualization solution that satisfy the service-level agreement (SLA) requirements. First, we apply an auto-scaling mechanism of route server in the database system. The experimental results show that the average response time of auto-scaling DB solution and none-scaling DB solution are 4.3 seconds and 7.1 seconds, respectively. Second, in order to minimize the impact when moving data to a new VM, we also design a shard data migration algorithm for the database system. The auto-scaling DB solution uses the algorithm to determine how many VM should be added and which data should be moved to those added VM.