ECHO - Enabling Cloud Hosted Organisations

This project proposes an investigation into the use of software for the automated deployment and management of heterogeneous resources as utility resource clouds based on previously funded EPSRC activity. It offers automated multi-provider cloud management and will bring an establish field tested technology with commercial features such as cost and budget setting, service level agreements, robustness and resilience to researchers and research applications.The proposal focuses on answering two questions:1. Can the cloud utility model replace local compute clusters or groups of clusters and provide an efficient, extensible and customisable computation platform for the majority of the research community?2. Can the utility cloud model bring a more flexible and dynamic compute infrastructure to researchers and free them of the need to spend time and money supporting their infrastructure?The project focuses on developing the EPSRC investment in Zeel/i technology to manage application access and deployment to cloud providers-the work will focus on supporting research applications rather than the commercial applications that have been the focus for the last 5 years.The pilot project will focus on two representative research application areas that provide distinctly different challenges for cloud as a research computation platform and where PI/co-PI expertise is available. Exemplar 1 is a suite of numerical mathematical applications which relies heavily on BLAS libraries. This application suite is often run in small and medium scale to test algorithm implementation but is designed for very large scale production runs when matrix sizes of 100k by 100k are normal, and where efficient BLAS implementation, multicore resources and fast storage are essential to give satisfactory execution times. Exemplar 2 is a suite of applications to support gene analysis and relies on a several large application databases and a collection of online remote databases and analysis services.The plan of work focuses on the execution, management and analysis of cloud based implementations of the exemplar applications on commercial utility providers-these are judged to offer the most diverse cloud environments and offer distinct cost savings when compared to owned and managed infrastructures. This approach also enables researchers to access the technology and cost savings that the rapid adoption of cloud within the commercial domain is bringing.

Researchers will benefit because * they can own infrastructure only for the period of time they require it; * they can request and use only the infrastructure they need for the application and the context of its use; * they can request an infrastructure that is suited and tailored to their application; * they are freed from the cost and effort involved in deployment, managing, supporting and refreshing owned infrastructure; * they can take advantage of the innovation that is occurring in the utility resource market aimed at commercial applications, such as resilience, mobility and scaling of applications; * they can use a wide variety of utility resource providers and technologies using a single application centric API; and * they can define a cost and budget to their application. These factors should reduce research costs, provide a flexible and efficient research infrastructure and enable researchers to focus on their research and supporting applications and not supporting the compute infrastructure necessary to enable their research and thereby enabling more ambitious research.