Support for Urgent Computing Based on Resource Virtualization

Abstract

Abstract. Virtualization technologies provide flexible execution environments that could bring important benefits for computational problems with strong deadlines. Large Grid infrastructures are becoming available nowadays and they could be a suitable environment to run such on-demand computations that might be used in decision-making processes. For these computation, we encounter the need to deliver as much resources as possible at particular times. These resources may be provided by different institutions belonging to a grid infrastructure but there are two important issues that must be satisfied. Firstly, all resources must be correctly configured and all the components needed by the application must be properly installed. If there is something small missing that is required then applications will fail. Secondly, the execution of urgent applications must be made quickly in order to produce useful results in time. If applications must wait in a queue, results might be useless because they are obtained too late. To address these issues, we describe a job management service, based on virtualization techniques, that avoids configuration problems and increases the number of available resources to run applications with critical deadlines. We describe the main components of our service that can be used on top of common batch queue systems and we show some experimental results that prove the benefits of applying time-sharing techniques on the virtual machines to increase the performance of urgent computations.

A Development and Execution Environment for Early Warning Systems for Natural Disasters

“…The number of such virtual worker nodes can be increased or decreased to best meet current needs. An example of such a system is the Virtual Spaces Job Manager [11]. Job Description Language extensions allow users to specify the required virtual machines and level of urgency.…”

Architecture of Environmental Risk Modelling: for a faster and more robust response to natural disasters

“…Indeed, the multiplicity of modelling dimensions (states; controls; uncertainty-driven arrays of parameters and scenarios; arrays of D-TM modules to account for software uncertainty) may easily lead to an exponential increase of the required computational processes (the so called "curse of dimensionality"). A viable mitigation strategy might be offered by HPC tools (such as Urgent HPC [79,80,81]) in order to sample highdimensional modelling space with a proper method.…”