Towards grid/clouds challenges for autonomic computing

Abstract

At the second Grids Meets Autonomics Computing workshop, held alongside the IEEE International Conference on Autonomic Computing (ICAC 2010), the panel will address the following issue: "Towards grid/clouds challenges for Autonomic Computing". The general objective of this panel is to define practical path (or paths) toward building bridges between the operational needs of production grids and clouds, and the solutions proposed by the Autonomic Computing community. The goal is to follow-up the issue raised at the GMAC'09 panel, and also at the ICAC'09 panel: how can AC demonstrate its contribution to addressing key issues in Grid and Cloud computing. The ideal outcome would be a plan for designing a (or a set of) Grid/CloudToAC "challenges". In this technical sense, a challenge is a self-contained problem, typically defined by datasets and objectives. Usually, a challenge is published through a web site (which also collects solutions), is widely advertised to the relevant scientific communities, and has a deadline. A ranking of proposed solutions is afterwards established and published; in some cases, financial incentives for the winner(s) are proposed. To which extent is this general scheme adequate to the above mentioned goal is by no means obvious, otherwise it would have been done before. However, gathering top-level actors from the two banks (to pursue the bridge metaphor) may be the key. A most important question lies in a careful consideration of the final evaluation step: for many questions of real interest in grids and clouds, there is no obvious "gold standard", that is a reference and quantifiable interpretation (analogue of labels in classification). Moreover, the challenge has to be well defined and limited (a general question such as "improve the reliability of the system" is not acceptable). Finally, setting up a challenge is technically demanding (web site, applications and submissions management, etc.). As it would be in-between the two communities, both would have to contribute. The panel will address the following questions. What would be the main advantages and the main issues for creating such Grid/CloudToAC challenges? What focus for these challenges and what would be the envisioned timescale for each of them? Which actions would be required to facilitate the actual realization of the selected challenges? This panel will bring four speakers contributing by their expertise in grid operational and design issues, autonomics, and engineering. The format is an exploratory Panel aimed at bootraping a dynamics in which the attendees will be invited to contribute. © 2010 Copyright is held by the author/owner(s).