Software Aging Analysis and Prediction Using AKNN Algorithm

Abstract

Distributed server's are used in every IT Firm and all the online-local and global communication today is greatly dependent on them. Therefore these servers undergo continuous usage round the clock, As a result of which performance of these servers degrades resulting to software aging. Software aging is the phenomenon that affects the performance of a software system drastically, thereby degrading it when functioning in a long running state. This is generally caused by factors like exhaustion or inappropriate use of system resources, the accumulation of internal errors and so on. In this paper, we are predicting software aging so that necessary precautionary measures can be taken before the server performance is actually affected. This work demonstrates a prediction model by taking NASA server log dataset information as an input. By the use of Modified Apriori and KNN Algorithm (AKNN) together, we are able to identify the number of anomalies successfully. This module fetches necessary information which includes: all the Host ID's that are facing anomaly, along with their respective failure percentages. Greater the failure rates more are the servers prone to aging. Results are discussed using graphical representations. The graphical representations include the sessions queried, Latency, total session duration and finally the number of anomalies detected with respect to computational time. Hence the proposed work combines two algorithms, and successfully predicts aging in a server with better accuracy and faster computational time.