Growth in Datacentre Electricity Use 2005 2010 Growth

Growth in Datacentre Electricity Use 2005 – 2010

Growth in Datacentre Electricity Use 2005 to 2010 2011 paper by Jonathan G. Koomey, Ph. D. and Consulting Professor, Stanford University. Main Topic • • Growth in Datacentre Electricity Use 2005 -2010 Results of growth in electricity use in datacentres from 2005 to 2010 Growth in electricity use vs. previous 5 year period Percentage of electricity used by datacentres - Globally & in the U. S. Reasons for the slowdown in growth Secondary Topic EPA Growth Projections for 2005 -2010 • Projected growth figures vs. actual growth • Reasons for differences between papers • Methods used to analyse data

Main Findings 1. Growth of Datacentre Electricity Use 2005 – 2010 • Globally - electricity use in datacentres grew by 56% • U. S. - electricity use in datacentres grew by 36% Both global and U. S. rates had doubled for 2000 - 2005 • Difference of 44% - Globally • Difference of 64% - U. S. Due to reduction in server installed base & changes to the power use per server

Main Findings 2. Percentage of Electricity Use 2005 – 2010 • 1. 1 - 1. 5% of global electricity use • 1. 7 – 2. 2% of U. S. electricity use Average Power Consumption • Global – 23. 2 to 34. 4 gigawatts • U. S. – 7. 7 to 12. 3 gigawatts

Main Findings 3. Projected Growth vs. Actual Growth • Globally - Electricity use projected to grow to 2. 2% of total electricity used vs. actual growth to 1. 3% • U. S. - Electricity use projected to grow to 3. 5% of total electricity used vs. actual growth to 2% Projected growth = All trends continue. Upper bound = IDC figures for no. of servers + ‘ 00 – ’ 05 electricity use per server continues. Lower bound = IDC figures for no. of servers + no change to electricity use per server. Actual growth = Mid-point of upper and lower bound

Reasons for Slowdown in Datacentre Electricity Use Virtualisation Reduction in number of servers • Higher server utilisation • Multiple virtual servers on single physical server • Less servers = Less electricity use Other Factors • Cloud computing • Datacentre efficiency • Power consumption Financial Crisis • Economic slowdown – companies cut costs by reducing numbers of servers • Less servers = Less electricity use per server

Breakdown of Electricity Use in a Datacentre Servers Infrastructure Communication Equipment Storage Devices Volume Servers Cooling Systems Switches Hard Disk Drives Mid-Range Servers Fans Hubs Solid State Disks High End Servers Pumps Routers UPS Battery Backups • Volume servers are the single biggest consumers of electricity in the datacentre • Future power per server more important than server installed base • Infrastructure equipment is the biggest group consumer of electricity • Storage capacity less important than number of spindles

Breakdown of Electricity Use in a Datacentre PUE - Power Usage Effectiveness: Used to measure efficiency of datacentre. Total Power Entering Facility Power Consumed by IT Equipment PUE Ideal PUE = 1. This means that all the power entering a datacentre is consumed by the IT equipment.

Why is this an area of interest to the electronics industry? Drivers to Reduce Electricity Use Political & Legal – legislation, regulation and policies Economic – Cut costs & increase profits Scientific – Prevent damage to environment (e. g. greenhouse effect) Social – Reputation of company, social influence businesses, corporate social responsibility Factors Affecting Future Electricity Use in Datacentres • • Virtualisation Cloud computing – Higher server utilisation & less electricity Power consumption per server vs. numbers of servers Datacentre efficiency – since 2006 datacentre efficiency improving through practices (e. g. more efficient heating & AC, smart lighting , etc. )

Questions?