Load balancing method for data management using high availability distributed clusters
Furuya Shota, Kiyoshi Ueda · 2017
Recently, users for various information and communication technology services have been increased with the advancement of terminals and the development of network infrastructure. Hence, systems using distributed data store are becoming widespread to improve scalability, fault tolerance and high reliability. For example, MAGONIA®server architecture, a part of NetroSphere Concept handles large-scale traffic by load-balancing. Basic functions include data redundancy for improvement of the reliability as well as system management in a dynamic scale. In previous studies, distributed data stores supported retrieval of attribute values by inverted index, but there are cases where bias to the index occurred. In this research, we propose a method to divide the hash value of an index in an application conforming to MAGONIA to equalize the load. In addition, a method to determine the bit length to divide is also shown. Then, we evaluate applicability of these proposed methods by actual machine and simulation.