Tuesday, February 23, 2010

CloudScape II

Day one of CloudScape II in Brussels, organised by OGF-Europe, was filled with much discussion. Following the first CloudScape, held last year, the afternoon was dedicated to use case scenarios for cloud computing featuring speakers from industry and research.
It kicked off with a talk from Paul Strong of eBay who championed the idea of companies being able to 'run their business from a cellphone'. In the long term, cloud could enable us to do this, allowing businesses to focus on their ideas (their major asset) rather than the technologies that implement them. But so far clouds are being leveraged mainly in small and medium enterprises - integrating clouds into large enterprises is a non-trivial task. This is something that the RESERVOIR project are looking at, giving a large-scale perspective of cloud technologies.

In science, clouds are being used by enterprising individuals to reduce the time from idea to realisation. Paul Watson gave a great example of a PhD student at his university who independently ran his experiment on cloud rather than using university resources. Paul described a collaboration between chemists and computational scientists at Newcastle University, UK who are using the cloud to aid drug discovery. In this area of research researchers often want to answer the question 'What is the activity of a given molecule?' While this can be determined through experimentation, it takes a long time, so scientists would like to be able to predict the answer. Chemists use a method called QSAR (Quantitative Structure Activity Relationship) to do so, using the properties of other molecules to make predictions for unknown molecules. However integrating new data into these models is computationally intensive – it could take 5 years to introduce a large data set into the model. The collaboration are seeing if they can speed this up by moving the models onto the Azure cloud, and using it to split and queue tasks before returning the results.

CloudScape is an opportunity for the community to discuss the challenges we face when trying to implement clouds in research and business. Although they can offer a number of benefits, clouds do not make it easier to build complex scalable dependable distributed systems. Davide Solomon from INFN stressed the importance of leveraging on the successes we've already achieved with grid. Grids have a long history with multi-domain large scale resource sharing and using the know-how and infrastructure we've created in grids could help with issues such as interoperability, open standards and privacy in clouds. This sentiment has been echoed by the Commission who stress the importance of protecting the investment we've made in grids so far while ensuring we have enough space to innovate with other technologies. INFN are working on the coexistance of grids and clouds together with a project called WNoDes. WNoDes is reusing existing know-how to integrate cloud services through grid. For example they are giving alternative access to their service (e.g. through Shibboleth) so cloud and grid services can all be accessed in the same way.

At the networking drinks after day one of the conference, I asked delegates how they thought this event compared to last year's. While one participant thought there was not much progress from last year, others though that we've moved from asking 'are clouds useful?' to 'how can we use them?' What everyone does seem to agree on is that the development of clouds is going to be hard work. The community will have to use all the skills they've learnt so far in areas such as grid and HPC. As Kyriakos Baxevanidis said in his presentation this morning it's likely that the strong interactions between EGI, which will soon begin, and cloud initiatives in the coming months will define cloud development in the future.

No comments: