Google Research

Precentile-Based Approach to Forecasting Workload Growth

  • Alex Gilgur
  • Stephen Gunn
  • Douglas Browning
  • Xiaojun Di
  • Wei Chen
  • Rajesh Krishnaswamy
IT Performance and Capacity by CMG 41st International Conference (CMG2015), Computer Measurement Group, 3501 Route 42 Suite 130 #121 Turnersville, NJ 08012-1734 USA

Abstract

When forecasting resource workloads (traffic, CPU load, memory usage, etc.), we often extrapolate from the upper percentiles of data distributions. This works very well when the resource is far enough from its saturation point. However, when the resource utilization gets closer to the workload-carrying capacity of the resource, upper percentiles level off (the phenomenon is colloquially known as flat-topping or clipping), leading to underpredictions of future workload and potentially to undersized resources. This paper explains the phenomenon and proposes a new approach that can be used for making useful forecasts of workload when historical data for the forecast are collected from a resource approaching saturation.

Learn more about how we do research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work