Optimal Design of LEMoNet for Environmental Monitoring of Data Centers

• Published in: IEEE transactions on green communications and networking Vol. 5; no. 4; pp. 1820 - 1832
• Main Authors: Jafarizadeh, Mehdi, Zheng, Rong
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.12.2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Analytical models; Bluetooth; Communication systems; Computer centers; Cooling; Data centers; Design optimization; Design parameters; Environmental monitoring; Green data centers; Internet of Things; Logic gates; Mathematical models; monitoring networks; Multiple objective analysis; performance evaluation; Power consumption; Sensors; Sustainable development; Temperature sensors; Topology; Wireless communication; Wireless networks; Wireless sensor networks
• ISSN: 2473-2400; 2473-2400
• DOI: 10.1109/TGCN.2021.3100592

Half of the electrical power in today's data centers (DCs) is consumed by cooling units, but much is wasted due to over-cooled or under-utilized servers. Environmental monitoring using Data Center Wireless Sensor Networks (DCWSNs) plays a central role in detecting and mitigating hotspots or over-cooling conditions. These networks often include thousands of sensors and require a reliable and power-efficient communication protocol. The Low Energy Monitoring Network (LEMoNet) is a two-tier DCWSN protocol that features Bluetooth Low Energy (BLE) for sensor communication in the first tier and leverages multi-gateway packet receptions in its second tier. This research studies the performance of LEMoNet in different DC topologies through the development of an analytical model. The model solves a multi-objective optimization problem to calculate two design parameters. The first parameter is the optimal number of gateways and an optimal location for each one. The second parameter is an optimal transition (TX) power for each sensor, subject to the application requirements. Evaluation results show that the optimal design includes one gateway on every other row. Also, the model reduces 65% of the sensors' power consumption via TX power optimization while Packet Reception Rate (PRR) is kept up at the level of 95%.