Operating systems include many heuristic algorithms designed to improve overall storage performance and throughput. Because such heuristics cannot work well for all conditions and workloads, system designers resorted to exposing numerous tunable parameters to users—thus burdening users with continually optimizing their own storage systems and applications. Storage systems are usually responsible for most latency in I/O-heavy applications, so even a small latency improvement can be significant. Machine learning (ML) techniques promise to learn patterns, generalize from them, and enable optimal solutions that adapt to changing workloads. We propose that ML solutions become a first-class component in OSs and replace manual heuristics to optimize storage systems dynamically. In this article, we describe our proposed ML architecture, called KML. We developed a prototype KML architecture and applied it to two case studies: optimizing readahead and NFS read-size values. Our experiments show that KML consumes less than 4 KB of dynamic kernel memory, has a CPU overhead smaller than 0.2%, and yet can learn patterns and improve I/O throughput by as much as 2.3× and 15× for two case studies—even for complex, never-seen-before, concurrently running mixed workloads on different storage devices.

Abstract Navigation is the process by which people control their movement in virtual environments and is a corefunctional requirement for all virtual environment (VE) applications. Users require the ability to move, controllingorientation, direction of movement and speed, in order to achieve a particular goal within a VE. Navigation israrely the end point in itself (which is typically interaction with the visual representations of data) but applicationsoften place a high demand on navigation skills, which in turn means that a high level of support for navigationis required from the application. On desktop systems navigation in non‐immersive systems is usually supportedthrough the usual hardware devices of mouse and keyboard. Previous work by the authors shows that many usersexperience frustration when trying to perform even simple navigation tasks — users complain about getting lost,becoming disorientated and finding the interface `difficult to use'. In this paper we report on work in progressin exploiting natural language processing (NLP) technology to support navigation in non‐immersive virtualenvironments. A multi‐modal system has been developed which supports a range of high‐level (spoken) navigationcommands and indications are that spoken dialogue interaction is an effective alternative to mouse and keyboardinteraction for many tasks. We conclude that multi‐modal interaction, combining technologies such as NLP withmouse and keyboard may offer the most effective interaction with VEs and identify a number of areas where furtherwork is necessary. ACM CSS: I.3.6 Computer Graphics Methodology and Techniques—Interaction and Techniques, I.3.7 Three‐DimensionalGraphics and Realism—Virtual Reality, I.2.7 Natural Language Processing—Speech Recognitionand Synthesis

Abstract Whilst providing images of excellent quality, ray tracing is a computationally intensive task. The first part of this paper compares the speed‐up achieved in ray tracing using various space subdivision algorithms and discusses the implications of implementing the algorithms on parallel processing systems. The second part addresses the problem of building the data structure within the rendering process, a situation which occurs when the rendering process is parallelised and dynamic scenes are rendered. Greater performance can be achieved with dynamic structure building compared to creation of the structure prior to rendering. The dynamic building algorithm proposed reduces the building time and storage cost of space subdivision structures, and decreases the data structure creation‐render cycle time, thus enhancing image parallelism performance.