Dynamic Right-Sizing for Power-Proportional Data Centers

By Minghong Lin, Adam Wierman, L. Andrew, Eno Thereska. IEEE/ACM Transactions on Networking, 2011

770 citations recorded by the source corpus.

Power consumption imposes a significant cost for data centers implementing cloud services, yet much of that power is used to maintain excess service capacity during periods of low load. This paper investigates how much can be saved by dynamically “right-sizing” the data center by turning off servers during such periods and how to achieve that saving via an online algorithm. We propose a very general model and prove that the optimal offline algorithm for dynamic right-sizing has a simple structure when viewed in reverse time, and this structure is exploited to develop a new “lazy” online algorithm, which is proven to be 3-competitive. We validate the algorithm using traces from two real data-center workloads and show that significant cost savings are possible. Additionally, we contrast this new algorithm with the more traditional approach of receding horizon control.

Subfields: Computer Networks and Communications, Information Systems

View Paper: https://www.semanticscholar.org/paper/686ea1426a50b5e263c1acb845d3d6f7558e9eb4

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Information Systems

Citations in the corpus

Toward Edge Intelligence: Multiaccess Edge Computing for 5G and Internet of Things (2020, background): “The operation of slowing down or switching off the operating speeds of some edge servers, which have light computational work-loads, is referred to as dynamic right sizing for realizing energy-proportional servers in literature [160].”
Greening geographical load balancing (2011, background): “Engineering advances in cooling, virtualization, multi-core servers, DC power, etc. have led to signi.”
A Survey on Mobile Edge Computing: The Communication Perspective (2017, background): “2) Geographical Load Balancing for MEC: GLB is another key technique for green DCNs [189], [190], which leverages the spatial diversities of the workload patterns, temperatures, and electricity prices, to make workload routing decision among different data centers.”
Online Optimal Control with Linear Dynamics and Predictions: Algorithms and Regret Analysis (2019, background): “…i) is able to capture a commonly observed phenomenon in predictions that short-term predictions are usually much more accurate than the long-term predictions; ii) allows researchers to derive insights for the role of prediction and possibly to extend to more complicated cases [31, 30, 43, 44].”
Understand Dynamic Regret with Switching Cost for Online Decision Making (2019, background): “e switching cost measures the dierence between two successive decisions, which is needed in many practical scenarios such as service management in electric power network [35], dynamic resource management in data centers [31, 33, 40].”
Stochastic models of load balancing and scheduling in cloud computing clusters (2012, background): “There are many design issues associated with such systems [7], [8], [9], [10], [11], [12].”
Online algorithms for geographical load balancing (2012, background): “This can be much tighter than the competitive ratio of 3 obtained by schemes unaware of future loads [17].”
Data Center Energy Consumption Modeling: A Survey (2016, methodology): “Such predictions can then be used to improve the energy efficiency of the data center, for example by incorporating the model into techniques such as temperature or energy aware scheduling [18], dynamic voltage frequency scaling (DVFS) [19][20][21], resource virtualization [22], improving the algori”
Thinking fast and slow: Optimization decomposition across timescales (2017, background): “To show this, we prove below that (1) can be reformulated as an online convex optimization problem, which is a classical online algorithms problem that has received considerable attention in the last decade [1], [9], [10], [25], [26].”
Renewable and cooling aware workload management for sustainable data centers (2012, background): “Categories and Subject Descriptors C.0 [Computer Systems Organization]: General Keywords sustainable data center, renewable energy, demand shaping, scheduling, cooling optimization 1.”
It's not easy being green (2012, background): “However, a datacenter can be made power-proportional by shutting off unused servers [28] and switches [18].”
Online Learning for Offloading and Autoscaling in Energy Harvesting Mobile Edge Computing (2017, background): “evices are hard and dangerous to reach. However,designinggreen MEC powered by renewable energy is much more challenging compared to green communication systems [13]–[15] or green data center networks [16] [7], [17], [18] since the radio and computation resources have to be jointly managed, whereas p”
On Global Electricity Usage of Communication Technology: Trends to 2030 (2015, background): “Measures to handle and reduce electricity usage and CO2 emissions from data centers have been widely reported [62,147–156].”
Resource Management in Clouds: Survey and Research Challenges (2015, background): “Work on the data center right sizing problem has continued from both algorithm development and system implementations perspectives; good examples of recent efforts are [48, 73, 106, 142].”