Feature - Virtual infrastructure

With origins tracing back to 1293, the Universidad Complutense de Madrid is one of the oldest universities in the world, and home of OpenNebula.

A problem of scale (and motley software)

Grid site administrators have numerous problems to deal with when configuring and scaling their infrastructure, including a large number of users with different software components, and physical resources that cannot be isolated or partitioned. Together, these can make for limited service and reliability, along with inflexibility.

To overcome these challenges, Ignacio M. Llorente and his colleagues of the dsa-research group at the Universidad Complutense de Madrid developed a new, open-source technology which acts as a virtualization layer between the service and the physical infrastructure layers, integrating existing grid and cluster middleware stacks.

It extends the benefits of  Virtual Machine Monitors (VMMs) from a single physical resource to a cluster of resources, decoupling the server not only from its physical infrastructure but also from its physical location.

Called OpenNebula, the technology works by dynamically overlaying VMMs over physical resources. Consequently, computing services such as SGE, Condor and OpenPBS can be executed on top of the virtual infrastructure, allowing a physical cluster to run multiple virtual clusters.

Not a cloud, but a nebula. Illustration courtesy of Ignacio M. Llorente

Benefits

    This separation of resource provisioning, managed by OpenNebula, provides:
•    Cluster consolidation. Because multiple virtual working nodes can run on a single physical resource, the number of physical systems and space, administration, power and cooling requirements are reduced. The allocation of physical resources to virtual nodes could be dynamic, depending on its computing demands, by taking advantage of the migration functionality provided by existing VMMs.
•    Cluster partitioning. The physical resources of a cluster can be used to execute virtual working nodes bound to different virtual clusters.
•    Support for heterogeneous workloads. Jobs with multiple or conflicting software requirements, can be run, even if they usually require a specific version of a library, or a legacy application.

Consequently, this approach provides the necessary flexibility to allow Grid sites to isolate and partition physical resources.

The idea of a virtual infrastructure to dynamically manage the execution of VMMs on physical resources is not new. Examples include Platform VM Orchestrator, IBM Virtualization Manager, Novell ZENworks, VMware Virtual Center and HP VMManager, among others.

The team says their system is different in that it offers high modularity and open architecture. The OpenNebula Engine provides a command line interface for monitoring and controlling VMs and physical resources, allowing integration with third-party tools such as local resource managers,  service adapters and VM image managers. In addition, they say that decoupling the service layer from the infrastructure layer allows a straightforward extension to any kind of service.

OpenNebula is a part of Reservoir, the European initiative in cloud computing.

—Ignacio M. Llorente, Universidad Complutense de Madrid