Hadoop on the EGI Federated Cloud

Diego Scardaci (EGI.eu/INFN)
With the rapid increase of data volumes in scientific computations, the importance of utilising parallel and distributed computing paradigms in data processing is becoming more and more important. Hadoop is an open source implementation of the MapReduce framework supporting processing large datasets in parallel and on multiple nodes in a reliable and fault-tolerant manner. Scientific workflow systems and science gateways are high level environments to facilitate the development, orchestration and execution of complex experiments from a user-friendly graphical user interface. Integrating MapReduce/Hadoop with such workflow systems and science gateways enables scientists to conduct complex data intensive experiments utilising the power of the MapReduce paradigm from the convenience provided by science gateway frameworks. This presentation and demonstration will illustrate how easily Hadoop clusters can be deployed on EGI FedCloud resources, Hadoop applications can be executed on these clusters, and finally resources can be released after execution. Users of the EGI FedCloud WS-PGRADE gateway can import and parameterise pre-prepared workflows for the above tasks published in a public workflow repository. Users can set the type/flavour and number of desired nodes in the Hadoop cluster, select the target EGI FedCloud site, and define the Hadoop executable and the desired data source and destination. All three functionality (create Hadoop cluster, execute Hadoop job, destroy Hadoop cluster) can be executed as a standalone job or can be combined into more complex workflows automating different user scenarios. The presentation and demo will be continued with an open discussion of use cases that could benefit from the presented setup. Connection link: http://connect.ct.infn.it/egi-inspire-vt/
Hadoop-EGI user manual v1.3
    • 10:00 10:05
      Welcome 5m
      Speaker: Diego Scardaci (EGI.eu/INFN)
    • 10:05 10:30
      Hadoop on EGI (talk) 25m
      Speaker: Tamas Kiss (University of Westminster, London, UK)
    • 10:30 10:45
      Hadoop on EGI (demo) 15m
      Speaker: Carlos Blanco (University of Cantabria)
    • 10:45 11:30
      Q&A and discussion of use cases 45m