Towards Multi-level Adaptation for Distributed Operating Systems and Applications

Abstract

Abstract. Distributed operating systems simplify building and executing applications on large-scale infrastructures, such as clusters, grids and clouds. These systems operate in a constantly changing environment characterized by varying application needs and varying physical infrastructure capabilities. To handle the diversity and dynamism of both the applications and the underlying infrastructures, the distributed Operating System (OS) should continually adapt to its changing environment. Two challenges arise in this context: how to design the distributed OS in order to facilitate dynamic adaptation, and how to ensure that OS-level adaptation does not conflict with application-level adaptation. This paper proposes to address these challenges by: (1) building the distributed OS as an assembly of adaptable services following the service-oriented architecture; and (2) using a common multi-level adaptation framework to adapt both the OS and the application layers in a coordinated way. Moreover, the paper presents experimental evidence of the usefulness of this approach in adapting the distributed shared memory service of a specific distributed OS.

Cross - layer monitoring and adaptation of multi - cloud service - based applications

“…Finally, in [Dib et al 2012] the authors make two contributions regarding dynamic adaptability of modern operating systems . First, they propose a serviceoriented approach for building distributed OSs, which enables adapting the OS algorithms in a dynamic, on-demand fashion.…”