Live big data analytics resource management techniques in fog computing for tele-health applications

• Published in: Jordanian journal of computers and information technology (Online) Vol. 7; no. 1; pp. 89 - 103
• Main Authors: Shihab, Raja, Muhammad, Huda K., Tahir, Muhammad
• Format: Journal Article
• Language: Arabic; English
• Published: Amman, Jordan Princess Sumaya University for Technology 01.03.2021; Scientific Research Support Fund of Jordan Princess Sumaya University for Technology; Scientific Research Support Fund of Jordan (SRSF) and Princess Sumaya University for Technology (PSUT)
• Subjects: analytics accuracy; apache hadoop2 yarn schedulers; Big Data; big data analytics; Data analysis; fog infrastructural management; fog/edge computing; patient monitoring; per-user control; smart hospital; stream analytics
• ISSN: 2413-9351; 2415-1076
• DOI: 10.5455/jjcit.71-1605864596

Enhancing the IoT health monitoring systems used in various environments, such as smart homes and smart hospitals, imply lively analyzing the patients’ critical streams (e.g. ECG stream). Conducting these tele-health applications over the traditional cloud violates the deadline constrains of the stream analytics applications, which results not only in performance degradation, but also in inaccurate analytics results due to patient's stream loss. Fog computing can take place within the patient's vicinity and is considered as the best candidate for critically analyzed stream applications. Fog nodes are geo-distributed and are poor in resources, thus a scalable and fault-tolerant resource management platform for stream analytics in fog computing is a must. Current Stream Processing (SP) resource managers are designed for massive resource nodes, deploying them over the poor resource edge fog nodes greatly decreasing the fog infrastructure utilization. Innovative SP resource managers that cope with the fog nature are needed. We propose Fog Assisted Resource Management (FARM) platform based on Apache Hadoop2 resource manager (YARN) for compatible stream/batch analytics. Static FARM (S-FARM) represents two YARN schedulers; per-user and per-module. Results indicate that per- user scheduler overcomes the lack of resources issues of the edge fog nodes, fully utilizes the fog infrastructure and allows the system to expand safely up to its double size. In addition, Differentiated S-FARM scheduler is proposed to support per-user control to the analytic results' accuracy and speed. Stream CardioVascular Disease (S-CVD) application for patient's ECG analytics is simulated in iFogSim to judge the proposed YARN schedulers. The research is pioneer in enhancing the poor resource edge fog node utilization, supporting per- user control to live big data analytics IoT applications and utilizing iFogSim to implement and evaluate the resource manager performance of a stream analytics platform.