Application workload pattern and its implications for resource provisioning in cloud: An early experience

Cloud computing is a flexible computing model where resources are allocated and deal located dynamically. The dynamic nature of allocating resources provide scope for optimizing the resources utilization. However the key challenge is to optimize the resource utilization without impacting the applications. In this context we argue that knowing the usage (behavior) of applications is important and using this information we can predict the future requirements of resources and plan the capacity accordingly. In this paper we describe our experiences with production level cloud applications and their usage pattern discovered through a clustering algorithm. Our technique is a completely automated technique for usage pattern discovery. We also share preliminary results on how these usage patterns can be related to resource utilization and help in planning resource allocation.