Chapter 19Allocating and Pricing Data Center Resources with Power-Aware Combinatorial Auctions

BENJAMIN LUBIN and DAVID C. PARKES

19.1 Introduction

As data centers become ever more essential to not only our economy but also our daily lives, it is increasingly important for them to be well managed. Three specific aspects are of paramount importance: ensuring that the right resources be allocated to the right use, ensuring that power is used only when it is providing real benefit, and making sure that the prices charged to users for service are providing incentives that support such allocative and power efficiency. In this chapter, we look at a method for employing a combinatorial market mechanism to achieve these ends.

In 2006, US data centers used about 61 billion kWh; that is, 1.5% of the 4 trillion kWh consumed in total. This is the amount of energy used by 5.8 million average US households (5% of all households) [1]. Producing this power resulted in 37 million metric tons of COc19-math-0001, or 0.6% of the 5.9 billion metric tons released from all sources. That is roughly 16% of that produced by the burning of jet fuel and more than that used to power TVs. This electricity cost $4.5 billion and required a peak load capacity of about 7 GW, more than double the level of consumption in 2000 [2, 3]. This pace of rising data center energy consumption has slowed, with global data center power ...

Get Smart Data Pricing now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.