The Virtual Machine (VM) Scaler: An Infrastructure Manager Supporting Environmental Modeling on IaaS Clouds

Abstract

Infrastructure-as-a-service (IaaS) clouds provide a new medium for deployment of environmental modeling applications. Harnessing advancements in virtualization, IaaS clouds can provide dynamic scalable infrastructure to better support scientific modeling computational demands. Providing scientific modeling "as-a-service" requires dynamic scaling of server infrastructure to adapt to changing user workloads. This paper presents the Virtual Machine (VM) Scaler, an autonomic resource manager for IaaS Clouds. We have developed VM-Scaler, a REST/JSON-based web services application which supports infrastructure provisioning and management to support scientific modeling for the Cloud Services Innovation Platform (CSIP) [Lloyd et al. 2012]. VM-Scaler harnesses the Amazon Elastic Compute Cloud (EC2) application programming interface to support model- service scalability, cloud management, and infrastructure configuration for supporting modeling workloads. VM-Scaler provides "cloud control" while abstracting the underlying IaaS cloud from the end user. VM-Scaler is extensible to support any EC2 compatible cloud and currently supports the Amazon public cloud and Eucalyptus private clouds versions 3.1 and 3.3. VM-Scaler provides a platform to improve scientific model deployment by supporting experimentation with: hot spot detection schemes, VM management and placement approaches, and model job scheduling/proxy services.