Log Analysis as a way to assist Opera mini cluster management decisions

Abstract

This thesis considers ways that analysis of Opera mini logs can assist decisions related to global and local load balancing of Opera mini clusters. The analy- sis is aimed to determine the distribution of traffic with respect to country of origin and server within the cluster over the period of 2 weeks by creating a system for extraction and analysis of log data. Findings show that a large part of traffic originates in Russia with India and Indonesia being second and third. This trend is expected to continue. The total number of requests is shown to be increasing exponentially. The internal load balancing algorithm shows un- even results due to memory effect and high load on the cluster during the peak time.