Joint Computing and Electric Systems Optimization for Green Datacenters

Abstract

This chapter presents an optimization framework to manage green datacenters using multi-level energy reduction techniques in a joint approach. A green datacenter exploits renewable energy sources and active Uninterruptible Power Supply (UPS) units to reduce the energy intake from the grid while improving its Quality of Service (QoS). At server level, the state-of-the-art correlation-aware Virtual Machines (VMs) consolidation technique allows to maximize server's energy efficiency. At system level, heterogeneous Energy Storage Systems (ESS) replace standard UPSs while a dedicated optimization strategy aims at maximizing the lifetime of the battery banks and to reduce the energy bill, considering the load of the servers. Results demonstrate, under different number of VMs in the system, up to 11.6% energy savings, 10.4% improvement of QoS compared to existing correlation-aware VM allocation schemes for datacenters and up to 96% electricity bill savings. IntroductionEver increasing demands for computing and growing number of clusters and servers in datacenters have ramped up the power consumption costs as an undesirable effect [21]. On the other hand, traditional fossil fuel concerns, carbon emissions and global warming impose the introduction of more sustainable energy sources and behavioural change of people [41], since 10% of the global consumption of electrical energy has been estimated to be consumed by Information Technology (IT) infrastructures [14].

Job Scheduling in High Performance Computing

“…Although there is a great promotion in using renewable energy, renewable energy only contributed 19.3% to humans' global energy consumption in 2017. Therefore, in recent years, HPC and data center facilities try to cut their traditional energy consumptions [23]- [28]. Therefore, there is a research area focusing on maximizing resource consumptions from renewable energy so as to reduce energy consumption from traditional energy sources.…”

ECOGreen: Electricity Cost Optimization for Green Datacenters in Emerging Power Markets

“…2-System). We use the green datacenter model proposed in our previous work [40] where datacenter equipment, PV modules, EES, and power grid are connected via a bidirectional CTI bus. In the following subsections we first present the used EES model, which is essential for correctly assessing the battery lifetime.…”

“…The EES comprises one homogeneous battery bank managed in a hierarchical fashion. The battery model is based on Peukert's law [41], and has been conceived as a plug-and-play component that can be easily replaced and adapted [40]. To correctly account for the benefits of the EES system we need to control the charging sequences of the battery bank, and to model battery ageing (i.e., power loss in EES).…”

“…Therefore, we choose the lithium-ion technology instead, which offers at least one order of magnitude higher number of cycles (useful for tracking fast time-changing RS signal), but at a higher cost. To maximize the lifetime of the storage and make the battery bank work in the optimal range of SoC, we constraint the maximum depth of discharge (DoD) to 70% [40]. This indicates that only 30% of the total capacity of the battery can be used for bidding problem.…”

“…The cell temperature (T j ) is obtained using Eq. 6, where T amb is the environmental temperature, G T = 800W /m 2 @ 20 • C and NOCT = 45.5 • C. We also tune the PV module size (the number of cells and panels) considering the peak power of the green datacenter with respect to its load, which is the most common approach to PV sizing [40].…”