Resources

Energy Management for Data Centers

Customer Story

Introduction

A large consumer electronics technology company sought out SevOne for granular infrastructure visibility at worldwide scale to address their growing data center energy and operational efficiency needs.

The Data Center Energy Efficiency Problem

Data centers have challenges apart from specific individual business challenges. There exists a vicious circle between compute capability, power consumption and cooling. Growing businesses are under constant pressure to add data center compute capacity (or open a new data center). Each unit of added compute capacity increases power consumption which creates thermal load, necessitating more cooling which again adds to power consumption. Lawrence Berkeley Laboratories has shown that the compute function of the data center accounts for 51% of power consumption and HVAC (cooling) accounts for 34%, with lighting and other functions consuming the remainder. Therefore 85% of the energy usage in an average data center is subject to regular growth and in need of efficiency measures. In small installations or for small businesses, this growth may be tenable, but for growing companies and at large scale across multiple data centers, the expense and complexity of data center infrastructure management has led to a strong need to drive energy and operational efficiency to keep the vicious circle in check.

Business Challenge

In this use case, the company faced this data center energy efficiency challenge with the extenuating circumstances of rapid growth in computing needs and a corporate desire to use renewable “green” sources of energy, which come at a premium cost to traditional energy. With over 600,000 servers and 25,000 smart power strips and power distribution units across multiple datacenters, the company had limited insight into:

  • Compute power consumption
  • Opportunities for server virtualization and server consolidation to reduce power consumption and cooling
  • Potential operation efficiencies through better utilization of available server resources

To gain the next level of insight into their infrastructure performance, the company turned to SevOne to implement monitoring, baselining and alerting for power consumption across their world-wide data centers.

Why SevOne?

SevOne was chosen for this company’s key long-term infrastructure initiative based on SevOne’s ability to monitor all types of infrastructure at near real-time with highly granular visibility combined with the scale to collect energy efficiency data from over 25,000 smart meters without loss of monitoring or reporting performance. SevOne extends to any infrastructure with the capability to integrate and correlate any time series data with any other performance monitoring data with automatic baselining, thresholding and alerting capabilities. As a relevant example, SevOne can combine and correlate data center temperature data with power consumption.

Business Results and Benefits

The first step in the plan to increase data center efficiency through infrastructure performance monitoring enabled the company to more accurately forecast data center needs and reign in unnecessary expansion. Just as classic network performance management enables IT administrators to forecast bandwidth capacity requirements, the infrastructure performance monitoring supplied by SevOne enabled the company to baseline power and cooling consumption alongside server usage to forecast future data center needs. SevOne’s automated baselining helped the company understand normal power consumption and ferret-out unusual conditions, even across the immense scale of the data centers. The company has postponed or cancelled several data center developments and envisions a 10-40% power savings as a result of understanding power consumption at the rack level.

These results, however, are really only the “low hanging fruit” of the power strip-based infrastructure monitoring program. The baseline and regular power performance data are being combined with IPMI data (CPU, Fans, etc.) collected by SevOne, to provide a more complete picture of power usage at the server level. This level of data will enable the company to improve efficiencies in a number of important ways:

  • Drive server providers to deliver improved power consumption
  • Engage vendors on new models of service
  • Develop compute efficiencies by server type and utilization
  • Plan for peak power utilization and adjust operations accordingly
  • Shift resource usage based on known use profiles

Potential for Expansion

The many remaining options listed are, themselves, just the beginning of the possibilities that the company can develop using SevOne to monitor data center infrastructure. The addition of new metrics and the potential for script-driven automation based on alerts and usage patterns supports ever increasing refinements in operations and improvements in efficiency:

  • Monitor data center temperatures and humidity by aisle or area enabling most efficient resource use based on cooling requirements
  • Monitor external weather and correlate to power usage to avoid spikes in power usage or unnecessary cooling
  • Use alerts and script controls to limit power consumption to avoid running resources at or near capacity
  • Monitor HVAC performance directly to drive efficiency

Summary

The seemingly small-scale effects of server power consumption become massive expenses at scale and require innovative infrastructure management to drive energy and operational efficiency. SevOne’s unique ability to monitor all types of infrastructure at near real time with highly granular visibility, and the capability to scale without loss of monitoring or reporting performance has enabled a large consumer electronics technology company to monitor and forecast power consumption and cooling in their data centers, driving energy and operational efficiencies. SevOne sees the future as the need to manage infrastructure, not just networks and is partnering with visionary companies to create a vibrant communications infrastructure.