Cluster Recovery for Fault Tolerance of Spatial Database Cluster in Sensor Networks

摘要

In sensor networks, a huge amount of data is collected by millions of sensors and small mobile devices need to be processed fast. And database which stores those data always should be able to response for any requirement. Spatial database cluster provides high performance and high availability. That suits sensor networks because spatial database cluster can efficiently manage and process much amount of data. The previous system, however, should write external logs every transaction for high availability in all nodes. So, all of update transactions are slowly processed because of writing external logs. Also, recovery time of the failed node is increased because external logs for all of database are written in only single storage. In this paper, we propose the cluster recovery of spatial database cluster. The proposed method has cluster logs in each table unit for consistency among nodes. Also, the cluster log is written just in case any node is failed. Therefore the proposed method is processed more fast because all update transactions don't write cluster logs. And the proposed method provides fast recovery because each table can be recovered concurrently.