A Design of Application-Aware Profiling and Monitoring Architecture for Cloud Computing

Abstract

A cloud computing system should put more attentions on system monitoring and application profiling because it demands better performance guarantee. For example, the cloud computing resource providers may charge users according to how the resources are used. Thus service quality implies business income. Previous grid computing approaches only focus on monitoring hardware components, failing to accurately depict the triangle relationship among the users, applications, and the resource providers. In this paper, we present an application-aware profiling and monitoring architecture which consists of three major components: the application-aware profiling agents, a profiling database, and filters. The application-aware profiling agents are installed on every computing node to keep track of the execution status of the applications. The observed information is then sent to the filters for preliminary processing and then to the profiling database, or directly to the profiling database. Observed data sets are formatted in the XML standard. Details of application activities thus can be found on the profiling database though user interfaces. The proposed architecture is able to integrate the information of users, applications, jobs, processes, and resources. When problems arise, applications with high performance guarantee can be identified to get timely service. This approach provides better performance guarantee and resource utilization. Based on our design, cloud computing service providers can develop different billing strategies, service-level agreements, and user privileges according to user payment and the observed application activity information.