An Architecture for Edge Computing over Underutilized Gateways
The Internet of Things (IoT) is growing at a rapid pace; it is projected that there will be more than 41 billion connected devices by 2027, an almost fivefold increase from 2019. To realize the full potential of these things, it is essential for IoT systems to facilitate rich, low-latency, and robust interactions between them. However, today’s IoT focuses more on the things than the Internet, with manufacturers confining their devices to vendored silos and limiting device-to-device interactions. Although this siloed, cloud-first design offers convenience for manufacturers, it suffers from two vital drawbacks. First, applications scale poorly when using cloud APIs to interact with devices that have different interfaces. Second, storing user data in the cloud leads to potential privacy and security risks, and executing applications on the cloud could lead to delayed response times. To address these issues, our research aims at creating a platform that enables applications to execute on the edge, and provides rich abstractions for device interaction, while offering better data privacy, security, and application latency to users.
We hypothesize that the typically underutilized gateways in IoT systems can be exploited as a computing platform to run applications on the edge rather than on the cloud. Applications can interact with devices connected to any gateway using standard abstractions provided by the platform. We envision that our platform would enable IoT applications to be run locally without cloud support, and provide seamless device interactions by allowing easy integration and standardized application abstractions. This work specifically highlights three key design challenges for the platform and seeks to find an answer to this question: can we address the identified design challenges to enable a cloud-like user experience on the edge, but with reduced response time and adequate scalability, while also guaranteeing privacy of data? The proposed solution is presented, and the platform is evaluated for scalability, as well as compared with other edge computing platforms.
Yuan Tian (Chair)
Brad Campbell (Advisor)