Delay Tolerant Computing
Mohammad Aazam, Khaled A. Harras, Ali E. Elgazar · 2018
The digital world is expanding rapidly and advances in networking technologies such as wireless broadband (WiBro), low-power wide area networks (LPWANs), 4G/5G, LiFi, and so on, are paving the way for the emergence of sophisticated services. The number of online and mobile applications leveraging sophisticated smart devices are increasing in complexity requiring more computation and communication. While current smart phones and other IoT devices are becoming more powerful, the support gap is widening when compared to the demand of current and future compute-intensive tasks such as those often required for smart health care, ambient assisted living (AAL), virtual/augmented reality, intelligent vehicular communication, and so on. For many of these applications, computational or data storage tasks can not be entirely performed locally and have thus exploited different offloading techniques. The focus in such techniques has traditionally been on minimizing delay when supporting such applications. In this paper, we argue for the opportunity to tap into idle compute resources to support another dimension of applications characterized with high task demand, but can tolerate larger delays. We present delay tolerant computing (DTC) paradigm, and define where it falls within the compute ecosystem. We highlight the potential behind DTC with various applications, and propose a generalized architecture for how this paradigm can be realized. Finally, we show, using a couple of case studies, the potential behind DTC when compared to other compute paradigms by providing preliminary experimental results based on monetary cost, computational requirements, and delay.