Till umu.se

Research Areas in Cloud Computing

By combining virtualization, distributed systems, and autonomic computing technologies, self-managed multi-cloud infrastructures for large-scale provisioning of compute or data intensive services can be made reality. These infrastructures give unprecedented gains in terms of performance, cost-efficiency, flexibility, robustness, and sustainability.

Research Topics

The research is outlined from the fundamental questions about how much and what type of resources to allocate and when and where to deploy them. Although the questions are easy to identify, they are intrinsically hard to answer due to extreme scale, complexity, and dynamics. In addition to addressing the core questions themselves, a lot of effort goes into various related questions, like understanding and redicting the workloads, providing knowledge for decision making, enabling dynamic service relocation, and many more. Target infrastructures for individual project span from individual servers to single mega-scale datacenters and large-scale distributed telecom clouds (unifying management of anything from mega-scale datacenters to mobile acess networks). Sample indivual topics are

  • Modeling of workload time-variations and footprints
  • Performance anomaly detection
  • Admission control for robust resource overbooking
  • Predictive analytics for management purposes
  • Proactive capacity autoscaling
  • Differentiated quality of service
  • Placement (scheduling) of virtual machines
  • Task and workflow scheduling
  • Live migration of large-scale virtual machines
  • Virtualization solutions beyond individual servers
  • Infrastructure simulations
  • Accounting and real-time usage quota enforcement
  • Distributed and hierarchical fairshare scheduling
  • Holistic systems management based on business level objectives

The research is performed in a range of collaborations including researchers with expertize on topics helping us to solve the problems, like control theory, mathematical statistics, optimization, performance engineering, etc, and problem owners like major datacenter and software industries. The list of explicit collaboration partners is long, but representative examples include IBM Haifa Research Lab, SAP (Belfast and Ra'anana), Lund University, Lawrence Berkeley Lab, Ericsson Research, Google, Intel, Royal Institute of Technology, Luleå University of Technology, Simula, Red Hat, National Technical University of Athens, University of Vienna, Imperial College London, Flexiant, Universidad Complutense de Madrid, Barcelona Supercomputer Center, and several more.

Sidansvarig: Erik Elmroth


Distributed Systems

Contact Information

Erik Elmroth
Department of Computing Science, Umeå universitet
SE-901 87 Umeå, Sweden 

Plan 4, MIT-huset, C431

Tel:  46 90 7866986

Mobil:  46 70 3153928

Fax:  46 90 58014