CLOUD

By Steve Fisher, Editor In Chief -- Last Week Platform Computing, announced that the European Bioinformatics Institute (EBI) selected Platform LSF MultiCluster to build an enterprise Grid solution. To dig a little deeper into the relationship between the two organizations and the technology involved, Supercomputing Online interviewed Peter Roberts, Platform’s Director of Sales for the UK and Petteri Jokinen, Head of Systems and Networking, EBI. SCO: Please tell the readers a bit about the recent agreement between Platform and EBI. ROBERTS: Platform particularly wanted to supply the solution for EBI as Platform sees the Life sciences area, and drugs discovery in particular to be hugely compute resource intensive. Platform has already supplied systems to the Sanger Institute, on the same campus as EBI. The mapping of the human gene presents opportunities never before possible for in-silico, virtual drugs discovery. However, we recognize that the post-genomic screening problem is enormous, with so much data and so many potential targets. With upwards of 1.7 billion compounds to screen, the only way forward is with a compute intensive approach using Grid technology. JOKINEN: We started using LSF about five years ago when we migrated from VMS to Unix. Back then we needed a queuing system that was needed to replace the VMS batch queuing system. The current computing infrastructure is based on Unix servers and Linux PCs. The recent agreement has enabled us to use the multiclustering feature and to have a substantially larger cluster. SCO: Please tell us about EBI's current computing resources. JOKINEN: We currently have: -above 10TB of centralized disk storage -pc farm has about 300 cpus -Compaq Alphaservers (80 cpus) -Sun servers (50cpus) -Silicon Graphics servers (total ~50cpus) -Gigabit LAN -miscellaneous workstations on desktops (>150) SCO: In your opinion, what advantages does grid computing have over a standalone supercomputer for this type of work. ROBERTS: Grid computing is granular in nature, therefore the "virtual supercomputer" which EBI has can grow as their requirements and their funding grows. Also, EBI do not have to make any difficult, long-term decision as to which computer manufacturer to go with, as grid technology from Platform allows them to add servers from any manufacturer in the future. SCO: What advantages do Platform's products have over competitive offerings? JOKINEN: Scalability, proven solution, interoperability between heterogeneous platforms. ROBERTS: The main areas are: -- Platform products have been tried and tested over the last 10 years so are functionally very rich. They can be installed and "just work" with very little fuss. This represents a very low risk proposition to EBI. Also, we have the largest critical mass of developers dedicated to distributed computing R&D, with approximately 140 developers (approx 50 with PhDs and MAs) -- Platform has a large UK subsidiary and therefore a proper infrastructure in the UK to install, support, provide consultancy. This is very important for such a production grid. The EBI grid is itself likely to be organic in nature, growing in ways not currently planned. -- Platform products work on heterogeneous clusters of servers from all manufacturers, covering all flavours of Unix and Linux. This was vital. SCO: What sort of an impact do you see this agreement and the resulting enterprise computing grid having on research? ROBERTS: There is a view that Grid computing is yet to arrive in a real production sense. This isn't true, Platform's 1500 customers worldwide prove that. EBI are one of the organizations showing that this is far from the truth. Grid technology works today. This will inevitably give other Pharms organizations large and small the confidence to build grids to accelerate drugs discovery. SCO: When completed, what will EBI's enterprise computing grid look like? What type of resources and capabilities are we talking about? JOKINEN: Only time will tell that. We are prepared for thousands of nodes. SCO: Platform and EBI will both be at this week's O'Reilly Bioinformatics conference. Anything planned for that show that you'd like to give us a "sneak peak" into? ROBERTS: We'll be announcing a new cluster management solution for Linux on Monday and also Yury Rozenman, our life sciences expert from Platform is speaking at the event about Grid computing in life sciences.