Dynamic Data Driven Cloud Systems for Cloud-Hosted CPS

摘要

A vast majority of modern applications and systems have taken advantage of cloud computing, however, a general lack of performance and reliability assurances from the cloud have deterred applications from the realm of cyber physical systems (CPS) and Internet of Things (IoT) that have stringent Quality of Service (QoS) requirements avail the cloud capabilities. To address these limitations, we propose dynamic data driven cloud systems (DDDCS). DDDCS is a new paradigm that learns and continuously enhances models of the cloud infrastructure itself by adaptively instrumenting it so that analyzing these models can steer the cloud infrastructure to utilize the right resource management strategies needed to host the QoS-sensitive CPS/IoT tasks in the cloud. This research explores the challenges in realizing the DDDCS paradigm and addresses them by developing cloud model learning and adaptive cloud resource management algorithms. DDDCS is being evaluated in the context of a variety of CPS/IoT applications, such as smart city transportation.