Le nuage égaré, 1937. Image courtesy Andre Kertesz estate/Sarah Morthland Gallery

Are grid-users interested in using cloud technologies?

Yes, says a recent survey by the StratusLab project. One online survey was conducted aimed at two different sets of people: grid system administrators and the other at grid users, with about 200 responses overall.

The survey results found that grids and clouds represent different and complimentary computing models that support e-science infrastructures. While grids such as the European Grid Infrastructure make up the physical structures of networked clusters of computers, clouds enable flexibility, customization and virtualized resources — features with the potential to improve the grid user experience.

A summary of the survey suggested that over two-thirds of grid-system administrators intend to use cloud technology within the next twelve months, and over three-quarters of grid users within the next 24 months.

Intriguingly, over one-third of both groups are already using cloud technologies. Popular examples include Dropbox, Amazon and Google cloud services. In terms of accessing cloud technology, a command-line interface was the preferred option for system administrators, while an application programming interface (API) was most useful for grid users.

The survey highlighted that users want to run a variety of applications in clouds, including straightforward sequential programs and parallel applications requiring synchronization between programs. Some algorithms require synchronization for parallel processing; this is because they need to split the computation so it can run simultaneously on multiple processors. Examples of this are Monte Carlo simulations which introduce unpredictability into nuclear physics models, and Navier-Stokes equations used for calculating the motion of fluids in engineering.

Tower Bridge, London Image courtesy telegramSam MM, stock.xchng

Why clouds?

The advantages of clouds are many:

• Clouds allow the separation of a software environment from a physical computer (virtualization), enabling multiple operating systems (OS) to run on one computer.

Alternatively, this arrangement allows for the migration of a 'virtual machine' from one physical computer to another; literally transferring the ‘mind’ of one machine into another. An example of this would be a computer running Windows as a host OS while simultaneously running a guest Linux OS with all its accompanying software applications.

• Clouds also reduce management costs, freeing up budgets for additional computing resources and taking up less hardware resources, thus being greener for the environment.

• Cloud computing could help to overcome conflicting software requirements. For example, researchers in the field of bioinformatics (the application of computer science to molecular biology) use many software packages to analyze data. It is not always easy to find grid resources with appropriate operating systems, memory or processor capacity. All this information could be packaged into a virtual machine and used on any grid infrastructure.

However, there is currently little support for integrating grids and clouds, and no uniform way of running a grid job on the cloud because the interfaces and APIs are specific to each technology. StratusLab aims to enable users to take advantage of cloud technologies via the grid.

This month, StratusLab released a “toolkit” based on OpenNebula, an open-source cloud-computing service that provides a way for grid system administrators to set up a cloud. “User feedback will be extremely valuable,” said Ruben Montero, technical coordinator of StratusLab.  
By this time next year it may be common place for scientific researchers to access cloud computing resources, but more work still lies ahead.

—Adrian Giordani, iSGTW. StratusLab was conceived In November 2008 as a collaboration of academic and industry partners to explore integration of cloud technologies and with grid services. In June 2010, the project began with European Union FP7 funding.