Cloud Computing for Scientific Simulation and High Performance Computing

Cloud Computing for Scientific Simulation and High Performance Computing

Adrian Jackson (Edinburgh Parallel Computing Centre, The University of Edinburgh, UK) and Michèle Weiland (Edinburgh Parallel Computing Centre, The University of Edinburgh, UK)
DOI: 10.4018/978-1-4666-2854-0.ch003


This chapter describes experiences using Cloud infrastructures for scientific computing, both for serial and parallel computing. Amazon’s High Performance Computing (HPC) Cloud computing resources were compared to traditional HPC resources to quantify performance as well as assessing the complexity and cost of using the Cloud. Furthermore, a shared Cloud infrastructure is compared to standard desktop resources for scientific simulations. Whilst this is only a small scale evaluation these Cloud offerings, it does allow some conclusions to be drawn, particularly that the Cloud can currently not match the parallel performance of dedicated HPC machines for large scale parallel programs but can match the serial performance of standard computing resources for serial and small scale parallel programs. Also, the shared Cloud infrastructure cannot match dedicated computing resources for low level benchmarks, although for an actual scientific code, performance is comparable.
Chapter Preview

2. Evaluation Of Cloud Infrastructures

For our investigations we used two different Cloud infrastructures; the Amazon Web Services Elastic Compute Cloud (Amazon EC2) and an academic Cloud infrastructure provided by the UK National Grid Service (NGS)1. We chose Amazon because, at this point in time, it is one of the largest commercial Cloud resource providers. It is also easy to gain access to; all a prospective user requires is a credit card and an Amazon account, and immediate access can be obtained. Amazon offer access to a range of resources (different sizes, operating systems, exclusive or non-exclusive use and so on) for different prices.

The NGS experimental Cloud service was the second infrastructure we used. Unlike Amazon, it can only provide non-exclusive access, thus allowing us to evaluate the impact that sharing the computational resources can have for the user. It also enabled us to directly compare an academic Cloud infrastructure to a commercial Cloud.

2.1. Benchmarks

The massive computational resources offered by (chiefly commercial) Cloud providers now offer a potential alternative to the specialist High Performance Computing (HPC) machines used for scientific simulation codes by academia and industry worldwide. Amazon, for instance, target this market explicitly by including HPC-specific images in their range of products. However, it is not immediately clear what performance these Cloud resources really offer, or what costs are associated with them, what functionalities they provide or how good their usability is, and how they compare with current HPC technologies in general.

Clouds for HPC or heavily computational loads offer exclusive access to computing hardware (i.e. only one virtual machine will be running on a processor or node). However, there are a range of Clouds that do not guarantee exclusive access to computational resources. In this situation, low-level benchmarks that evaluate the performance impact of this sharing of resources are essential to enable users to make an informed choice over the computing resources to use for their particular requirements.

Therefore, as previously mentioned, both the Amazon EC2 and the NGS Cloud were used for the performance benchmarks. The Amazon Cloud was used to evaluate the parallel functionality currently being offered (i.e. the performance going beyond a single image) and the NGS Cloud was used for the low-level benchmarks of shared Cloud resources.

Complete Chapter List

Search this Book: