DOI: 10.5176/978-981-08-7656-2 A-15
Authors: Shakti Mishra, D.K. Yadav, D.S.Kushwaha, A.K.Misra
Abstract: Utilization of idle capacities of the computing nodes
has seen continuous built of research interest over the past one
decade and the corporate world finds it even more promising.
Among the various delivery models being pursued by the
industry is the cluster computing model that may be trusted or
open access system. This paradigm strives to provide many of the
distributed computing benefits like high availability,
manageability, scalability and dependability. To make these
features persistent over long period of time, environment must be
reliable and fault tolerant enabling the computing nodes to join
and leave the cluster easily. Therefore, an efficient mechanism
for configuration of clusters to attain high availability, scalability
and fault tolerance is required. This paper extends the Jingle-
Mingle model for load balancing [1] to a B+-tree based
architectural framework of self organizing clusters to attain load
balancing while also ensuring failure recovery and fault
tolerance. We propose various operations for the proposed self
organizing clusters that ensure proper mix of under loaded and
overloaded nodes. It also updates the clusters according to their
churn ratio. The experimental results prove that the average cost
of network restructuring with our proposed framework is
considerably lower. We also observe that although the number of
nodes in clusters vary, the cost of load balancing operation is
O(1).
Keywords: Load Balancing; Clusters; Auto Configuration; B+-
tree.
Updating...