Exascale computing poses numerous challenges that require innovation in all facets of computing: hardware, software, and algorithms. While a criticality assessment of the innovations required to achieve Exascale is pending, experts agree that the single most important challenge is power: Exascale computing will not be feasible unless the power requirements of Exascale computing systems are dramatically reduced. Exascale systems will need to deliver performance three orders of magnitude higher than current Petascale systems.
However, the current projections of hardware technology indicate that Exascale systems will also require one order of magnitude more power than current Petascale systems: While a 2011 Petascale system requires 5--7 MegaWatt to operate, an Exascale system will require up to 200 MegaWatt, excluding the power needed for storage and cooling, which may add another 200 MegaWatt in the power budget. The cost of operation of current Petascale infrastructures already runs in the range of 5-10 Million Euros annually and current experience suggests that for every Watt of power consumed by computing equipment in large-scale computing infrastructures an additional Watt needs to be invested in cooling the equipment. While hardware technology advances towards more energy-efficient devices, solving the power challenge for Exascale needs holistic solutions, which must include power-efficient system software. A number of Ph.D. projects are available in this area. These are outlined in the following sections.
Energy-efficient runtime systems for parallel programming: This is a family of projects that aims to implement new methods for multi-component (cores, memory, interconnect) power scaling and for improving system energy-efficiency, within the language runtime system. These methods will be based on a unified resource management methodology that consolidates allocated system resources to maximize opportunities for power savings in software. The projects will demonstrate these methods in one or more state of the art parallel programming libraries and languages, such as MPI, OpenMP, Cilk and POSIX Threads. Furthermore, the projects will use emerging programming models for Exascale computing based on partitioned global address spaces and asynchronous dataflow execution.
Energy-aware memory management for Exascale systems: This project aims to implement new software methods and tools for reducing the energy footprint of memory hierarchies, which are rapidly becoming a dominant energy consumer of high-end computing systems. Heterogeneous memories provide a viable alternative to reduce the energy budget of memory hierarchies while sustaining performance. The project will develop new methods for managing heterogeneous memories for high-end computing systems. These include but are not limited to, heterogeneity-aware and power-aware memory allocation and data placement, memory device aggregation for low-power/high-throughput operation, and energy-aware explicit locality management.
A minimum of a 2.1 honours degree or equivalent in Computer Science or Electrical and Electronic Engineering, or relevant degree is required.
This 3 year PhD studentship, funded by the Department for Employment and Learning (DEL), covers approved tuition fees and a maintenance grant of approximately £13,590. The PhD studentship normally commences on 1 October 2012 however, this could be available immediately to suitable candidates.
Applicants should apply electronically through the Queen's online application portal at: https://dap.qub.ac.uk/portal/
Further information available at: http://www.qub.ac.uk/schools/eeecs/Scholarships/PostgraduateResearchScholarships/
|Supervisor Name:||Prof Dimitrios S. Nikolopoulos|
|Address:||School of Electronics,
and Computer Science
Bernard Crossland Building,
|Tel:||+44 28(0) 9097 1809|