Choose timezone
Your profile timezone:
Powered by Indico
v3.3.3-pre
Environmental awareness and rising energy prices have made energy-efficiency an important issue for managing computing centers and other IT systems. So far, research has mostly focused on hardware and infrastructure aspects, i.e. developing more efficient hardware or optimising cooling but studying the system as a whole has received less interest.
We study the optimization of computing systems by viewing a computing facility as a production unit on which we apply well established operations management principles to run it more efficiently. Our research goal is to find ways to improve throughput and energy-efficiency by scheduling requests based on their estimated properties and loads of all components (CPU, memory, network) using operations management principles of optimal loading.
Because of variations in jobs and the system behaviour, it is not possible to find a fixed optimum utilisation level. Therefore, we developed a fuzzy logic-based method that dynamically adapts parameters. In this way, it is possible to keep, for example, memory consumption stable with different workloads and this way achieving higher throughput and energy-efficiency than by using a traditional fixed number of jobs or fixed memory threshold approaches. Our preliminary results in scientific computing scenarios show that this approach can decrease energy consumption by 50% while increasing throughput up to 100% when compared to the current standard practices in scientific computing.
Prof. Ari-Pekka Hameri is a professor of Operations Management at the University of Lausanne and the director of the Technology program at the Helsinki Institute of Physics. He received his PhD in Industrial Management from the Helsinki University of Technology in 1993. His research has focused on efficient use of the full spectrum of production resources. He has published over 50 articles in international management and science journals concerning industrial IT, and management of production, projects and supply chains. He has been involved with numerous international research and spin-off projects related to industrial IT and production management.
Dr. Tapio Niemi is the leader of the GreenIT project at the Helsinki Institute of Physics (HIP). He received his PhD in Computer Science from the University of Tampere (2001). Prior joining HIP Technology Programme at CERN in 2004 he was a scientist and assistant professor at the University of Tampere. His research interests are green computing and data management. He has over 30 international publications in this field.
Organised by: Nils Hoimyr and Miguel Angel Marquina
Computing Seminars /IT Department