Multi-Tenant Virtual GPUs for Optimising Performance of a Financial Risk Application

Graphics Processing Units (GPUs) are becoming popular accelerators in modern High-Performance Computing (HPC) clusters. Installing GPUs on each node of the cluster is not efficient resulting in high costs and power consumption as well as underutilisation of the accelerator. The research reported in this paper is motivated towards the use of few physical GPUs by providing cluster nodes access to remote GPUs on-demand for a financial risk application. We hypothesise that sharing GPUs between several nodes, referred to as multi-tenancy, reduces the execution time and energy consumed by an application. Two data transfer modes between the CPU and the GPUs, namely concurrent and sequential, are explored. The key result from the experiments is that multi-tenancy with few physical GPUs using sequential data transfers lowers the execution time and the energy consumed, thereby improving the overall performance of the application.

On the Virtualization of CUDA Based GPU Remoting on ARM and X86 Machines in the GVirtuS Framework

The astonishing development of diverse and different hardware platforms is twofold: on one side, the challenge for the exascale performance for big data processing and management; on the other side, the mobile and embedded devices for data collection and human machine interaction. This drove to a highly hierarchical evolution of programming models. GVirtuS is the general virtualization system developed in 2009 and firstly introduced in 2010 enabling a completely transparent layer among GPUs and VMs. This paper shows the latest achievements and developments of GVirtuS, now supporting CUDA 6.5, memory management and scheduling. Thanks to the new and improved remoting capabilities, GVirtus now enables GPU sharing among physical and virtual machines based on x86 and ARM CPUs on local workstations,computing clusters and distributed cloud appliances.

Is Our Students Learning - QuestionMark: Extending E-Assessment Systems for Adapative Marking and Feedback

Our key contribution is a flexible, automated marking system that adds desirable functionality to existing E-Assessment systems. In our approach, any given E-Assessment system is relegated to a data-collection mechanism, whereas marking and the generation and distribution of personalised per-student feedback is handled separately by our own system. This allows content-rich Microsoft Word feedback documents to be generated and distributed to every student simultaneously according to a per-assessment schedule.

The feedback is adaptive in that it corresponds to the answers given by the student and provides guidance on where they may have gone wrong. It is not limited to simple multiple choice which are the most prescriptive question type offered by most E-Assessment Systems and as such most straightforward to mark consistently and provide individual per-alternative feedback strings. It is also better equipped to handle the use of mathematical symbols and images within the feedback documents which is more flexible than existing E-Assessment systems, which can only handle simple text strings.

As well as MCQs the system reliably and robustly handles Multiple Response, Text Matching and Numeric style questions in a more flexible manner than Questionmark: Perception and other E-Assessment Systems. It can also reliably handle multi-part questions where the response to an earlier question influences the answer to a later one and can adjust both scoring and feedback appropriately.

New question formats can be added at any time provided a corresponding marking method conforming to certain templates can also be programmed. Indeed, any question type for which a programmatic method of marking can be devised may be supported by our system. Furthermore, since the student’s response to each is question is marked programmatically, our system can be set to allow for minor deviations from the correct answer, and if appropriate award partial marks.

A parallel pattern for iterative stencil + reduce

We advocate the Loop-of-stencil-reduce pattern as a means of simplifying the implementation of data-parallel programs on heterogeneous multi-core platforms. Loop-of-stencil-reduce is general enough to subsume map, reduce, map-reduce, stencil, stencil-reduce, and, crucially, their usage in a loop in both data-parallel and streaming applications, or a combination of both. The pattern makes it possible to deploy a single stencil computation kernel on different GPUs. We discuss the implementation of Loop-of-stencil-reduce in FastFlow, a framework for the implementation of applications based on the parallel patterns. Experiments are presented to illustrate the use of Loop-of-stencil-reduce in developing data-parallel kernels running on heterogeneous systems.