ENEDI: Energy Saving in Datacenters

Abstract

Despite significant advances in the design and fabrication of power-efficient electronics and microprocessors, and the ubiquitous availability of Class A-rated lightning, cooling and heating appliances, modern datacenters account for a growing percentage of global energy consumption, currently estimated at 1.5%. The effort to reduce the environmental impact and carbon footprint of datacenter operations has led to the emergence of green datacenters designed to reduce energy consumption and/or use renewable energy to power computing and peripheral devices. One of the challenges that arise in the context of solar-powered datacenters is the intelligent adjustment of energy consumption to variations of solar energy production, changing environmental conditions (temperature, humidity, etc.), and fluctuations in computing demand. In this paper, we present the design and early implementation of ENEDI, an integrated system that collects, integrates and analyzes data from Internet-of-Things (IoT) sensors, operating system, as well as, cloud middleware monitors, and open-data sources regarding: i) photovoltaic PV energy production in solar-powered datacenter facilities

ii) weather conditions

iii) power consumption by datacenter subsystems (clusters, cluster nodes, and cooling equipment), and iv) the time-evolving profiles of containerized cloud applications running on the datacenter.