Pervasive Computing at the Edge

Paramvir Victor Bahl, Ramon Caceres, Nigel Davies, Roy Want
2020 IEEE pervasive computing  
Google & TODAY, THE INFRASTRUCTURE needed to support pervasive computing and the Internet of Things (IoT) is unparalleled as entirely new classes of applications and systems emerge. For example, pervasive systems designed to augment human cognition with tasks such as face recognition must operate at "superhuman speeds," delivering insights to help with human decision-making within very strict and narrow time limits. Similarly, the emergence of pervasive video analytics demands processing of
more » ... large volumes of video data in near-real-time. In general, the field of pervasive computing is rapidly changing in the face of major advances in sensing, data processing techniques, and wearable computing. The ever increasing data rates of high-speed networks also factor into the design tradeoff to decide if computing should be local, or remote. The design landscape is multifaceted and complex. Edge computing (for example, cloudlets and fog computing) offers the promise of supporting the demanding requirements of the next generation of pervasive computing and IoT systemsproviding devices with access to nearby processing and storage while simultaneously offering new solutions to issues such as user privacy and resilience. In this new world, research challenges abound: How should these systems be architected? What new applications are enabled? What are the challenges in deploying edge technology? And, how can they be secured against cyberattacks? In 2009, IEEE Pervasive Computing published one of the first papers on Edge Computing ("The Case for VM-Based Cloudlets in Mobile Computing") and in this special issue, we return to the topic to explore recent developments in the field. Our first article "OpenRTiST: End-to-End Benchmarking for Edge Computing" considers the challenge of comparing the performance of different edge-computing solutions. This is crucial to the future adoption of edge technologies as the approach requires deploying significant amounts of computing infrastructure-and the costs and benefits of these needs to be clearly articulated. The authors introduce OpenRTiST, an open-source application that is highly demanding in terms of computation, bandwidth, and latency requirements and thus is able to provide an ideal test for edge computing solutions. Our second article "Edge Computing for Legacy Applications" introduces a new use case for edge computing, namely supporting legacy
doi:10.1109/mprv.2020.3032205 fatcat:zn6735a6wjaztme4bazi2itx6u