The IRMOS/ISONI Real-Time Cloud Infrastructure: a Virtualised e-Learning Case-Study

Introduction Nowadays, applications are increasingly developed and made available in distributed infrastructures, where users access their services from remote. In the Cloud Computing model, applications are developed by Software-as-a-Service (SaaS) providers, by means of tools made available by Platform-as-a-Service (PaaS) providers, for being deployed over the resources made available by Infrastructure-as-a-Service (IaaS) providers. The viability of IaaS is dependent on the use of virtualisation technologies, which allow for deploying multiple virtual machines (VMs), hosting multiple Operating Systems and services, over the same physical hosts, achieving an increased server consolidation level. In this evolving scenario, more and more distributed applications with tight interactivity and timing requirements are being deployed over virtualised IaaS infrastructures. Unfortunately, when multiple VMs are deployed over the same physical resources (e.g., links and CPUs), the level of performance experienced by each VM is not stable any more, but it depends heavily on the overall workload imposed by the other VMs competing for the shared resources. However, using proper scheduling technologies and performance modelling techniques, it is possible to deploy virtualised distributed applications with a stable performance level, as being experimented with the virtualised Cloud Computing infrastructure developed in the IRMOS EU Project. In this paper, we show how these concepts have been practically applied to a real e-Learning application. A more extended discussion can be found in the original article [8] appeared at the IEEE SOCA 2010 Conference.