The Laundromat Model for Autonomic Cluster Computing

Traditional High Performance Computing systems require extensive management and suffer from security and configuration problems. This paper presents a new cluster-management system, called Evil Man, that aims at making clusters as secure and self-managing as possible. Evil Man is inspired by real-life Laundromats: All nodes in a cluster are configured with a minimal software base consisting of a Virtual Machine Monitor and a remote bootstrapping mechanism and customers then buy access using a simple pre-paid token scheme. All necessary application software, including the operating system, is provided by the customer as a full Virtual Machine and boot-strapped or migrated into the cluster Technically, the core ingredients of Evil Man are a novel operating system Boot-strapping and Self-Migration mechanism and a simple pre-paid Token Scheme on which all access control and accounting is based. Combined, these mechanisms create a simple, autonomous, secure and highly flexible cluster computing platform suitable for deployment in a cluster or Grid system. We work hard at reducing the amount of privileged network-facing software on each node. For instance, the privileged part of our network protocol implementation consists of only a few hundred lines of C-code. Performance measurements of our prototype show that Evil Man can bootstrap a non-trivial application onto an 11-node cluster in less than nine seconds and that performing live migration during application execution has negligible performance impact.