Author:

Keywords:

Science & Technology, Technology, Computer Science, Information Systems, Computer Science, Software Engineering, Computer Science, Theory & Methods, Computer Science, Internet of Things (IoT), Energy Harvesting, Benchmarking, Self Adaptive Networks, Industrial IoT Networks, POWER

Abstract:

© 2018 Association for Computing Machinery. The deployment of Internet of Things (IoT) devices is accelerating across a wide range of applications. The majority of today's IoT devices are powered by batteries that can operate for a maximum of a few years, after which they need to be replaced. This introduces two problems. First, the effort that is required to manually replace batteries cannot economically scale to support the next billion IoT devices. Secondly, treating billions of toxic batteries as disposable is not environmentally friendly. Together, these problems form a critical road-block in deploying IoT solutions. The biggest problem facing the designers of IoT applications is ensuring that their application software is energy efficient enough to operate within the strict power envelope that is provided by batteries or energy harvesting hardware. In this paper, we tackle this problem through the introduction of a distributed benchmarking middleware that rapidly and accurately quantifies the power consumption of different software configurations. Critically, our middleware operates in real-time across a distributed network of devices, allowing developers to experiment with code changes at runtime. This makes it significantly easier for developers to write applications that operate within the power constraints of batteries or energy harvesting systems. We evaluate our approach on a real world energy harvesting testbed and demonstrate that benchmarking results are accurate, with limited overhead for developers.