The phase 1 SKA is a massive radio telescope which will be deployed around 2020. Currently, teams from the member countries are creating the design of many components. The science data processor (SDP) is an approximately 300PF system which will process 2 exa-bytes per day on a continuous basis to extract science data products such as images from the signals. The pipeline of calculations requires many different kinds of parallelization and optimization each interwoven with others. To meet cost and power budgets extraordinary efficiency for each of the core algorithms has to be achieved, but computational efficiency is only a part of the story. We suspect this will be typical of many data intensive computing efforts of the future. Come to hear how this appears to be perfectly possible and what features of software and hardware we deem critical and desirable!
Dr. Peter Braam is a multidisciplinary innovator. Without fear to be wrong or incomplete, he advances technology frontiers and articulates insights in keynotes, white papers and private discussions. Peter is most attracted to grand challenges in technology and science, and research organizations engage with him to create concrete, focused visions and involve him in co-design of solutions. Peter is currently working with Cambridge University on the SKA telescope Science Data Processor and is best known as the creator of the widely used Lustre file system. Prior to founding and running five startups (the first four of which were acquired) related to parallel computing and programming languages, Peter was a senior academic at Oxford and Carnegie Mellon.