, Université Paris-Sud 11
, Cardiff University
Pages: pp. 9
This excerpt reports on a panel that took place in conjunction with the 2009 IEEE International Conference on Autonomic Computing. The full panel report, including the panelists' recommendations, is available for free on Computing Now (http://computingnow.computer.org/panel).
At the Grids Meet Autonomic Computing workshop that took place in conjunction with the IEEE International Conference on Autonomic Computing this year (ICAC 2009), we organized a panel called "Grids/Clouds/Autonomics Convergence." This panel focused on key research challenges in grid and cloud computing that autonomic computing techniques could support. The panelists were Marc-Elian Bégin of SixSq, Robert (Bob) Jones of CERN and director of the European Enabling Grids for E-sciencE ( EGEE) project, Manish Parashar of Rutgers University and the US National Science Foundation Center for Autonomic Computing, and Onn Shehory of IBM Haifa Labs. We asked them to consider these questions:
Because both grid and cloud computing infrastructures involve complex aggregation of resources, all panel members noted that more effective systems management is important. They observed that existing cloud computing infrastructure already supports some autonomic concepts, such as determining the number of virtual machines (VMs) to support, providing dynamically expandable storage, and migrating workloads across computational platforms. They also discussed the fact that the key differentiator between grids or clouds and enterprise data centers, from the viewpoint of the role of autonomics, is the nature of the applications and usage modes. Although existing autonomic computing research can directly be applied at the infrastructure level (server, VM, cluster management, and so on), it can't be applied at the application level (or at least wasn't being applied that way when the panel took place). The panelists also said that several types of cloud environments exist, so it's necessary to understand how to use autonomic concepts across multiple cloud owners and providers.
Autonomic computing techniques could address various aspects of system behavior. For instance, it could provide
Read more at http://computingnow.computer.org/panel.