• Publications
  • Influence
Dynamic right-sizing for power-proportional data centers
TLDR
We 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. Expand
  • 539
  • 55
  • PDF
Online algorithms for geographical load balancing
It has recently been proposed that Internet energy costs, both monetary and environmental, can be reduced by exploiting temporal variations and shifting processing to data centers located in regionsExpand
  • 245
  • 32
  • PDF
Greening geographical load balancing
TLDR
Energy expenditure has become a significant fraction of data center operating costs. Expand
  • 284
  • 25
Greening Geographical Load Balancing
TLDR
We show that if the price of electricity is proportional to the instantaneous fraction of the total energy that is brown, geographical load balancing significantly reduces brown energy use. Expand
  • 227
  • 23
  • PDF
Geographical load balancing with renewables
TLDR
We perform a trace-based study to evaluate three issues related to achieving this goal: the impact of geographical load balancing, the role of storage, and the optimal mix of renewables. Expand
  • 186
  • 13
  • PDF
Optimality, fairness, and robustness in speed scaling designs
TLDR
This work examines fundamental tradeoffs incurred by a speed scaler seeking to minimize the sum of expected response time and energy use per job. Expand
  • 133
  • 12
  • PDF
Dynamic Right-Sizing for Power-Proportional Data Centers
TLDR
Power consumption imposes a significant cost for cloud services, yet much of that power is used to maintain excess service capacity during periods of low load. Expand
  • 165
  • 8
A tale of two metrics: simultaneous bounds on competitiveness and regret
TLDR
We consider algorithms for “smoothed online convex optimization” (SOCO) problems, which are a hybrid between OCO and MTS, and show that no known algorithm can achieve sublinear regret and a constant CR, even when the objective functions are linear. Expand
  • 63
  • 5
  • PDF
Heavy-traffic analysis of mean response time under Shortest Remaining Processing Time
TLDR
We characterize the heavy-traffic growth rate of the mean response time under SRPT in the M/GI/1 system under general job size distributions. Expand
  • 25
  • 5
  • PDF
Joint optimization of overlapping phases in MapReduce
TLDR
This work studies the scheduling challenge that results from the overlapping of the "map" and "shuffle" phases in MapReduce. Expand
  • 78
  • 4
  • PDF