Design and Deployment of Interactive Petabyte Scale Applications

点击翻译按钮获取中文摘要

This proposed research program will establish fundamental properties for a platform designed to support a new class of applications in spatial computing. These data intensive applications must support highly collaborative environments with 3D emulation, manipulation and visualization of petabyte scale datasets in close to real-time. Comprehensive infrastructure for spatial computing has outstripped conventional computing platforms and requires integration not only with local and remote resources, but also must include peer--to--peer collaboration capability, and mixed reality interactive front-ends. This research will enable a new class of transformative applications that will offer unrivalled emulation experiences, deepening and diversifying the next generation of experts in any spatial domain. Small examples of this class of application have emerged, largely relying on local computing options with relatively small data sets. For example, medical practices are beginning to leverage advanced image processing and 3D technologies to create exact physical replicas of a patient's characteristics, allowing clinical teams to tailor care in unprecedented ways; environmental experts are beginning to fuse data from multiple sensors, satellites and other imagery to study local indications of climate change; and educational options are being enhanced through the creation of custom hands--on experiences for students using Augmented and Virtual Reality (AR/VR). But data sets are getting bigger------for example, new satellites will provide 20 TB of data per day. Currently however, there are no platforms for systematically deploying applications that capitalize on globally available, interdependent, petabyte scale data sets. Storage must be distributed, discoverable, and will require new open specifications that will evolve as the data sets continue to grow. The anticipated outcomes of this work will be threefold: (1) a methodology for building petabyte scale spatial computing applications based on a reference architecture derived from earth observation and remote sensing, (2) contributions to AI--assisted tool chains for development, deployment, emulation and monitoring of spatial applications, (3) design patterns of best practices for future applications incorporating tradeoffs in performance, scalability, availability and consistency. Students will work directly with stakeholders to evaluate the contributions of this proposed program, and our team will publish results for researchers and practitioners in both software engineering and systems. By targeting the fundamental tradeoffs at the platform level, this work will benefit researchers involved in any spatial computing domain. In terms of benefits on a national scale, we already working with stakeholders in the Data Commons projects within the Canadian Digital Supercluster.