Lee Chung Ho , Jang Yong Il , Bae Hae Yeong

The KIPS Transactions:PartD, Vol. 10, No. 6, pp. 935-948, Oct. 2003
10.3745/KIPSTD.2003.10.6.935,   PDF Download:

Abstract

An online scaling method adds new nodes to the shared-nothing database cluster and makes tables be reorganized while the system is running. The objective is to share the workload with many nodes and increase the capacity of cluster systems. The existing online scaling method, however, has two problems. One is the degradation of response time and transactions throughput due to the additional overhead of data transfer and replica's consistency. The other is an inefficient recovery mechanism in which the overall scaling transaction is aborted by a fault. These problems deteriorate the availability of a database cluster. This paper presents an advanced online scaling method that solves these problems and improves the availability of a shared-nothing database cluster. To avoid the additional overhead throughout the scaling period, our scalingmethod consists of two phases: a parallel data transfer phase and a combination phase. The parallel data transfer phase resources of spare nodes. Also, our method presents an efficient and fast recovery mechanism in the case of a fault during scaling period. Our experiments show that the proposed scaling method reduces the possibility of failure throughout the scaling period and improves the availability of the database cluster.

Statistics
Cite this article