Provisioning Virtual Clusters in the Cloud using Wrangler
SESSION: Research Poster Reception
EVENT TYPE: Poster
TIME: 5:15PM - 7:00PM
AUTHOR(S):Gideon Juve, Ewa Deelman
ABSTRACT: Cloud computing is becoming important for science applications. Unlike traditional cluster and grid platforms, clouds are not pre-configured to run scientific applications. Thus, tools need to be developed that can help users deploy their applications in the cloud. Our poster describes a system that automatically provisions and configures collections of cloud resources called virtual clusters. Users describe the cluster nodes and their desired roles using simple XML, and the system provisions and configures resources based on that description. Complex virtual clusters can be described using dependencies, which allow nodes to be dependent upon one another, and groups, which specify nodes that must be configured simultaneously. Dependencies and groups ensure that nodes are configured in the correct order and allow nodes to be dynamically added and removed from a virtual cluster at runtime. The system supports several different resource providers, and is designed to allow new providers to be added easily.