Article Title Questions

Cost-Benefit Analysis

of Cloud Computing versus Desktop Grids

Derrick Kondo1, Bahman Javadi1, Paul Malecot1, Franck Cappello1, David P. Anderson2

1INRIA, France,2UC Berkeley, USA

Contact author: derrick.kondo@inria.fr

Abstract

Cloud Computing has taken commercial computing by storm. However, adoption of cloud computing platforms and services by the scientific community is in its infancy as the performance and monetary cost-benefits for scientific applications are not perfectly clear. This is especially true for desktop grids (aka volunteer computing) applications. We compare and contrast the performance and monetary cost-benefits of clouds for desktop grid applications, ranging in computational size and storage. We address the

following questions: (i) What are the performance trade offs in using one platform over the other? (ii) What are the specific resource requirements and monetary costs of creating and deploying applications on each platform? (iii) In light of those monetary and performance cost-benefits, how do these platforms compare? (iv) Can cloud computing platforms be used in combination with desktop grids to improve cost-effectiveness even further? We examine those questions using performance measurements and monetary expenses of real desktop grids and the Amazon elastic compute cloud.

1 Introduction

Computational platforms have traditionally included clusters, and computational Grids. Recently, two cost-

efficient and powerful platforms have emerged, namely cloud and volunteer computing (aka desktop grids).

Cloud Computing has taken commercial computing by storm. Cloud computing platforms provide easy access to a company’s high-performance computing and storage infrastructure through web services. With cloud computing, the aim is to hide the complexity of IT infrastructure management from its users. At the same time, cloud computing platforms provide massive scalability, 99.999% reliability, high performance, and specifiable configurability. These capabilities are provided at relatively low costs compared to dedicated infrastructures. Volunteer Computing (VC) platforms are another cost efficient and powerful platform that use volunteered resources over the Internet. For over a decade, VC platforms have been one of the largest and most powerful distributed computing systems on the planet, offering a high return on investment for applications from a wide range of scientific domains (including computational biology, climate prediction, and high-energy physics). Since 2000, over 100 scientific publications (in the world’s most prestigious scien-

tific journals such as Science and Nature) [15, 5] have documented real scientific results achieved on this platform.

Adoption of cloud computing platforms and services by the scientific community is in its infancy as the performance and monetary cost-benefits for scientific applications are not perfectly clear. This is especially true for volunteer computing applications. In this paper, we compare and contrast the performance and monetary cost-benefits of clouds for volunteer computing applications, ranging in size and storage. We examine and answer the following questions:

• What are the performance trade-offs in using one platform over the other in terms platform construction, application deployment, compute rates, and completion times?

• What are the specific resource requirements and monetary costs of creating and deploying applications on

each platform?

• Given those performance and monetary cost-benefits, how do VC platforms compare with cloud platforms?

• Can cloud computing platforms be used in combination with VC systems to improve cost-effectiveness

even further?

To help answer these questions, we use server measurements and financial expenses collected from several

real VC projects, with emphasis on projects that use the BOINC [1] VC middleware. With this data, we use backof-the-envelope calculations based on current VC storage