Abstract - Datacenters are facing increasing pressure to cap their carbon footprints at low cost. Recent work has shown the significant environmental benefits of using renewable energy for datacenters by supply-following techniques (workload scheduling, geographical load balancing, etc.) However, all such prior work has only considered on-site renewable generation when numerous other options also exist, which may be superior to on-site renewables for many datacenters. Alternative ways for datacenters to incorporate renewable energy into their overall energy portfolio include: construction of or investment into offsite renewable farms at locations with more abundant renewable energy potential, indirect purchase of renewable energy through buying renewable energy certificates (RECs), purchase of renewable energy products such as power purchase agreements (PPAs) or through third-party renewable providers. We propose a general, optimization-based framework to minimize datacenter costs in the presence of different carbon footprint reduction goals, renewable energy characteristics, policies, utility tariff, and energy storage devices (ESDs). We expect that our work can help datacenter operators make informed decisions about sustainable, renewable-energy-powered IT system design.
The growth in the scale and number of datacenters is raising serious concerns about their power consumption. In 2005, the EPA  projected datacenter power demands to double by 2010. Though the recent Koomey report  has scaled down this growth to 56% (attributed to hardware improvements and increasing adoption of best practices), the cost of powering the U.S.'s datacenters is still expected to exceed $15 billion over the next decade and impose a peak load of over 20GW on the grid. Each 100MW power plant costs $60-100 million to build and emits over 50 million tons of CO2 over its lifespan . If datacenters were to be treated as a country, they would be the fifth largest electricity consumers across the world today .
Their high power consumption has two serious consequences for datacenters. First, generating and delivering this power to the datacenter, especially for its peak capacities and at times-of-day when there is high demand elsewhere, results in a high monthly electricity bill. A large datacenter may face millions of dollars annually in power-related operational expenditures. Second, much of the current grid power in many geographies is still heavily dependent on burning fossil fuels. Similar to other large consumers of power, datacenters find themselves increasingly pressured (either through legislation or simply public opinion) to find options to reduce their carbon footprint. Demand reduction is one obvious way of addressing both these concerns, and there have been numerous academic and commercial endeavors for achieving this with better energy-proportional computing technologies (consolidation and server shut down, deeper sleep states, and power mode control of IT equipment), improving power delivery efficiencies, and more efficiently controlling the cooling systems, over the past decade. Over and beyond demand reduction, datacenters are continuing to explore options for further reducing power related costs and their carbon footprints.
A complementary way of addressing these issues is with smarter electricity sourcing strategies. One is no longer necessarily tied to source from the grid. Capital costs of deploying renewable energy generation equipment (e.g., wind turbines, solar panels) have become increasingly attractive (especially with incentives in several geographies). These equipments could be deployed on-site (captive generation) at the datacenter facility itself, e.g., the Green House Data wind-powered datacenter  and Facebook's solar-powered datacenter . The advantages of such on-site generation include negligible transmission and distribution losses, and perhaps even the ability to tolerate an outage on the regular grid. However, it is not necessary that the best location for a datacenter (which can be a function of numerous other factors including network latencies, labor force availability, tax structures, etc.) necessarily has the right renewable energy potential for a profitable on-site renewable deployment.
Another model is to locate the renewable energy generation plant at an off-site facility (with good wind speed or solar irradiation), and "wheel" the generation across the grid to the consuming datacenter. In this model, along with transmission losses, there could be wheeling (and banking) charges imposed by the grid, though the generation potential may be much superior because of the flexibility to locate the generation in a more conducive location. In both these models, the mismatch between the production/supply and the consumption/demand may warrant consideration of explicit energy storage, and costs for this storage (either in explicitly procuring and managing storage devices such as batteries, or payment of banking charges to the grid) will need to be considered.
Whereas the above options require explicit involvement of the datacenter in provisioning renewable generation plants, there also exist a number of implicit options to achieve the same result. With one such set of options, a datacenter can purchase various renewable electricity products. One example is a power purchase agreement (PPA), buying a portion of the "green" power output from a renewable energy project in a long-term contract. Alternatively, a datacenter can simply procure its desired "blended" power mix from a third party provider at the applicable tariff. Such offerings may themselves come from a mix of "black" (i.e., fossil fuel based) and "green" (i.e., renewable) power sources - we refer to such a mix of black and green power sources as "brown." These renewable electricity products may be attractive since they eliminate the need for capital and operational investments for running renewable power plants, and perhaps also offer immunity to the variability inherent in renewable generation. Another set of implicit options is based on carbon offsetting, either through accredited CDM (Clean Development Mechanism) projects in developing countries or through purchase of carbon credits or renewable energy credits (RECs) in the open market. The merits of these implicit options, particularly the latter, are subject to the vagaries of a continuously evolving market.
Given all these choices, along with the vagaries of renewable energy generation capacity, variances in datacenter demand, and market price fluctuations, energy capacity planning becomes difficult. It is exactly this problem that this paper addresses by presenting an optimization framework to help a datacenter achieve a target carbon footprint at minimal cost.
We evaluate this optimization framework with a diverse range of datacenter power profiles, different procurement/offset mechanisms, and different kinds of generation efficiencies for both on-site and off-site renewables using traces from National Renewable Energy Laboratory (NREL) . Our evaluations show several interesting insights:
In this section, we provide background on datacenter power infrastructure and various options based on explicit or implicit incorporation of renewable energy for a datacenter to meet its carbon offsetting targets (if any) and/or cost-savings. Throughout the section, we follow up general concerns related to an aspect with specific assumptions or simplifications we make in our formulation.
A. Datacenter Power Infrastructure
Power enters the datacenter through a utility substation, which acts as its primary power source. Datacenters also employ diesel generators (DG) as a secondary backup power source. A typical datacenter power infrastructure consists of a hierarchy of power supply/distribution elements. Given our focus on decision-making related to renewable incorporation, rather than considering datacenter design completely from scratch, we focus on a datacenter that is already designed in the following sense: our datacenter's IT, cooling, and power infrastructure have already been provisioned based on well-established capacity planning techniques, but without employing any renewable energy options. Treating this datacenter as a given and a black box, we are interested in the subsequently arising capacity planning problem of choosing from among various explicit and implicit renewable energy options available to this datacenter. Our setting, therefore, captures an existing datacenter interested in altering its carbon footprint without the option of modifying its internal infrastructure. Studying the problem of joint capacity planning of the datacenter's IT, cooling, and power infrastructure with its renewable energy portfolio is part of our future work. Figure 1 captures this setting and shows different options for renewable incorporation.
B. Carbon Offsetting Targets
Many carbon cap policies and regulations are being deployed worldwide. They may be government-mandated, utility-imposed, or voluntary. For example, under European Union Emissions Trading System (EU ETS), the governments of the EU member nations agree on national emission caps. Large carbon emitters in these countries must monitor their CO2 emissions and report them annually to the government. Those who fail to offset or reduce their carbon footprints to comply with the carbon regulations face penalties. An alternative policy is based on the notion of a carbon tax, an environmental tax levied on corporate carbon footprints. As big power consumers, datacenters are facing increasing pressure to cap their carbon footprints. The life cycle carbon footprint of a datacenter includes the carbon emissions during the processes of IT equipment manufacturing and renewal (servers, UPS, cooling, etc.) and datacenter operation (which includes the electricity drawn from the utility).
Dr. Bhuvan Urgaonkar, PhD has over 15 years of experience in the field of Software Engineering and Computers. His work includes research in computer systems software, distributed computing (including systems such as Zookeeper, Redis, Memcached, Cassandra, Kafka), datacenters, cloud computing, storage systems, energy efficiency of computers and datacenters, big data (including systems such as Hadoop, Spark). He serves as an expert / technical consultant with multiple firms helping them (i) understand technical content related to state of the art products in areas such as content distribution, distributed computing, datacenter design, among others and (ii) interpret patents in these areas and connections between them and state of the art products and services. Services are available to law firms, government agencies, schools, firms / corporations, and hospitals.
©Copyright - All Rights Reserved
DO NOT REPRODUCE WITHOUT WRITTEN PERMISSION BY AUTHOR.