High Availability for Database Systems in Cloud Computing Environments

Maintaining availability in the face of hardware failures is an important goal for any database system. Users have come to expect 24x7 availability even for simple non-critical applications, and businesses can suffer costly and embarrassing disruptions when hardware fails. Many database systems are designed to continue serving user requests with little or no disruption even when hardware fails. However, this high availability comes at a high cost in terms of complex code in the database system, complex setup for the database administrator, and sometimes extra specialized hardware. In this talk, I will discuss approaches for providing high availability for database systems by taking advantage of the capabilities of cloud computing technologies. I will describe RemusDB, a system for providing active/stand-by replication for high availability based on full virtual machine state replication. I will also describe DBECS, a scalable, highly available, multi-tenant database service that is built on the Cassandra eventually consistent cloud storage system. DBECS carefully exploits the semantics of database transactions to provide strong transactional guarantees on top of the eventually consistent Cassandra, while retaining Cassandra's good scalability, low write latency, and partition tolerance, which we exploit for high availability. The goal of both of these projects is to take the complexity of high availability out of the database system, and let the cloud infrastructure below the DBMS provide high availability as a service.