Research And Utilize To The Database Middleware In Massive Data And High Availability Scenarios

The development of internet provides possibility for distributed data processing. However, the applying of distributed computing brought new challenges for managing and operating databases. It’s a great challenge to do the formation of traditional relational databases into a distributed system and meet the high performance, high availability, high scalability requirement at the same time.For a distributed-system, Percona Xtra DB Cluster provides consistency and availability, My SQL Replication provides availability and partitioning tolerance. In an e-commerce system, we need consistency, availability, partitioning tolerance in different scenarios. We proposed a new database cluster middleware called My Cat Plus to integrate the entire cluster and provide unified interface. My Cat Plus provides load balancing for SQL routing, it also satisfy read frequently with My SQL Proxy. We analysed the features of Hui Jin Shopping Mall System and proposed new methods to do sharding and caching. For scenarios that need high availability, we use Percona Xtra DB Cluster to meet this demand. It’s necessary to guarantee transaction as the data have been distributed across multiple nodes. We use two-phase commit to implement this feature. We implemented My Cat Plus based on My Cat in Hui Jin Shopping Mall System, and do experiments to compare with My SQL Cluster. The experiments show that My Cat Plus have advantages in performance and recovery speed.