Thumbnail Image

Energy-Aware Algorithms for Greening Internet-Scale Distributed Systems Using Renewables

Internet-scale Distributed Systems (IDSs) are large distributed systems that are comprised of hundreds of thousands of servers located in hundreds of data centers around the world. A canonical example of an IDS is a content delivery network (CDN) that delivers content to users from a large global deployment of servers around the world. IDSs consume large amounts of energy and their energy requirements are projected to increase significantly in the future. With carbon emissions from data centers increasing every year, use of renewables to power data centers is critical for the sustainability of data centers and for the environment. In this thesis we design energy-aware algorithms that leverage renewable sources of energy and study their potential to reduce brown energy consumption in IDSs. Firstly, we study the use of renewable solar energy to power IDS data centers. A net-zero IDS produces as much energy from renewables (green energy) as it needs to entirely off-set its energy consumption. We develop effective algorithms to help minimize the number of solar panels provisioned for net-zero IDSs. We empirically evaluate our algorithms using load traces from Akamai's global CDN and solar data from PVWatts. Our results show that for net-zero year, net-zero month, and net-zero week, our optimal algorithm can reduce the number of panels by 36%, 68%, and 82% respectively, thereby making sustainability of IDSs significantly more achievable. IDSs consume a significant amount of energy for cooling their infrastructure. Therefore, next, we study the potential benefits of using open air cooling (OAC) to reduce the energy usage as well as the capital costs incurred by an IDS for cooling. We develop an algorithm to incorporate OAC into the IDS architecture and empirically evaluate its efficacy using extensive work load traces from Akamai's global CDN and global weather data from NOAA. Our results show that by using OAC, a global IDS can extract a 51% cooling energy reduction during summers and a 92% reduction in the winter. Finally, we study the greening potential of combining two contrasting sources of renewable energy, namely solar energy and open air cooling (OAC). OAC involves the use of outside air to cool data centers if the weather outside is sufficiently cold and dry. Therefore OAC is likely to be abundant in colder weather and at night-time. In contrast, solar energy generation is correlated with sunny weather and day-time. Given their contrasting natures, we study whether synthesizing these two renewable sources of energy can yield complementary benefits. Given the intermittent nature of renewable energy, we use energy storage and load shifting to facilitate the use of green energy and study trade-offs in brown energy reduction based on key parameters like battery size, number of solar panels, and radius of load movement. We do a detailed cost analysis, including amortized cost savings as well as a break-even analysis for different energy prices. Our results show that we can significantly reduce brown energy consumption by about 55% to 59% just by combining the two technologies. We can increase our savings further to between 60% to 65% by adding load movement within a radius of 5000kms, and to between 73% to 89% by adding energy storage.
Research Projects
Organizational Units
Journal Issue
Publisher Version
Embedded videos