An Agent Based System for Monitoring and Measuring Performance of Public Cloud Applications for Users

Abstract

Monitoring of public cloud applications is a task of paramount importance for both Providers and consumers. The users of cloud and the providers are tied by service level agreements (SLA) that outline the provision and usage of the service. For the cloud application users’ application performance and minimum down times are key performance indicators of great concern in any SLA, noting that users of public cloud have no control over the cloud application. There are third party commercial monitoring tools used to monitor cloud application performance parameters mainly provided by the cloud providers. These tools run on the applications providers’ servers and measure performance from the applications side. The commercial tools are provider specific and may not present enough details for clients to be able to customize the behavior of the tools to accommodate their specific requirements. The tools may also not capture the users’ true experiences of the cloud application performance. This study proposes an agent based system that uses agents to interact with the cloud application and measure performance of the application as captured during the user agent interactions. The system measures performance as perceived by the end user hence reflecting the true user experience from the users’ side. Users’ perception of performance namely timeliness, availability and scalability were measured in this project. The researchers approach used user agents that interact with the cloud applications on behalf of the users and an intelligent monitoring agent that captures transaction durations during user interactions. In this project Prometheus Methodology of developing agent based systems was used for system specification, architectural design and detailed design. JADE framework was used in implementation of the system more specifically development of the agents. The results of the system indicated that agents were able to interact with the cloud application on behalf of the users and during the interactions the monitoring agent was able to capture and measure transaction durations of the cloud applications automatically and continuously updated the database. The transaction durations measured were interpreted as performance measures namely system availability, system scalability and application response time. These measures were displayed on a graphical interface and allowed users to access real time and historic data for monitoring of performance of cloud application.