Green Computing Green Computing Data centers extremely wasteful

Green Computing

Green Computing • Data centers extremely wasteful – Need energy to power – Need energy to cool • 1000 racks, 25, 000 sq ft, 10 MW for computing, 5 MW to dissipate heat • Need a system more efficient, less expensive strategy with immediate impact on energy consumption

Data Centers -2006/7 • Focus by green computing movement on data centers (SUVs of the tech world) • DOE Study started in 2006 (Bush years? ) – Don’t know how many data centers in US – 61 B k. Wh of energy – Cost: $4. 5 B (more than used by all color TVs in US) – In 2007, DOE reported data centers 1. 5% of all electricity in US – Greenhouse gas emission projected to more than double from 2007 to 2020

Goal • Within a few years, cost of power for data center was expected to exceed cost of original capital investment • Fed. Gov. wanted data center energy consumption to be reduced by at least 10% – Same as energy consumed by 1 M average US households

Servers - 2007 • One focus Servers – Rarely completely idle – Optimized at maximum utilization – Seldom operate at maximum – 10 -50% of max utilization levels – 100% utilization not acceptable for meeting throughput, etc. – no slack time

Servers - 2007 – Completely idle server waste of capital • Difficult to idle subset of servers – But Servers need to be available • Perform background tasks • Move data around • Can help recovery of crash – Applications can be restructured to create idle intervals • Difficult, hard to maintain – Devices with highest energy savings, highest wakeup penalty, e. g disk spin up

Disks: Inactive/active - 2007 • Penalty for transition to active from inactive state makes it less useful – Disk penalty 1000 higher for spin up than regular access latency – Only useful if idle for several minutes (rarely occurs) – More beneficial to have smaller penalty even if higher energy levels – Active energy savings schemes are useful even if higher penalty to transition because in low energy mode for longer periods

Suggested Solutions • Mismatch between servers’ high-energy efficiency characteristics and behavior • Design machines that consume energy in proportion to amount of work performed – No power when idle (easy) – Nearly no power when little work (harder) – More as activity increases (even harder)

Suggested Solutions • Sources of computing power in remote server warehouses • Located near renewable energy sources – wind, solar • Usage shifts across globe depending on where energy most abundant

Actualized Solutions • Processors close to energy-proportional – Consume < 1/3 power at low activity (70% of peak) – Power range less for other components • < 50% for DRAM, 25% for disk drives, 15% for network switches • Some “low hanging fruit” approaches – Orient racks of servers to exhaust in a uniform direction • Higher fruit - Microsoft – Built near hydroelectric power in WA – Built in Ireland - can air cool, 50% more energy efficient – Countries with favorable climates: Canada, Finland, Sweden and Switzerland

More Solutions • Google – trying to reduce carbon footprint Carbon footprint includes direct fuel use, purchased electricity and business travel, employee commuting, construction, server manufacturing – According to Google, its data centers use ½ industry’s average amount of power – How? Ultra efficient evaporative cooling (customized) • Green Grid consortium – Dell, IBM, Sun VM-Wear AMD • Green 500 – 500 most green supercomputers

More Solutions • US government – EPA has phase-one of Energy Star standards for servers – Measure server power supply efficiency and energy consumption while idle – Must also measure energy use at peak demand

Data Center Product Specification Completion – Energy Star • 2009 – Servers v 1. 0 • 2011 – Data Center Buildings Program • 2012 – UPS v 1. 0 (uninterruptable power supply) • 2013 – Servers 2. 0 – Storage v 1. 0 • 2014 – Large Network Equipment v 1. 0 • 2015 – Data Center Cooling Equipment v 1. 0

Where are we today?

2016 DOE Report • DOE 2008 report to Congress – First time data center use for the entire country was quantified in some way • 2016 report – surprised by how energy use curve flattened between 2008 and 2016 • Demand for data center capacity grew tremendously over last 5 years • Total data center energy grew only slightly

2016 DOEReport • 2000 -2005 data center energy consumption grew by 90% • 2005 -2010 energy consumption grew by 24% • 4% increase 2010 -2014 • US data centers consume 70 B kilo-watt hours in 2014 – 2% of total energy consumption, same consumed by 6. 4 M avg. US homes

Why? • Why? – Amount of servers not growing as quickly – Use virtualization and 2008 crash slowed growth • 3% annual increase in server shipments going to hyperscale data centers – cloud giants – Cloud giants created a science out of maximizing server utilization and data center efficiency – data center providers improvements in efficiency of facilities infrastructure, power, cooling equipment

Why? • Companies deploying fewer servers • Amount of power each server needs not growing as quickly • Power requirements static since 2005 • Servers better at reducing power when idle or at lower utilization • Data center power and cooling more efficient • Storage devices and networking hardware also significant efficiency improvements

One Solution - Energy proportional servers

• Efficiency improvements – • If at 2010 levels, data centers consume 40 B k. Wh more

Where are the data centers? • Enterprises host 71% of IT assets • 20% hosted by data center providers • 9% hosted in the cloud – Small data center: fewer than 40 enclosures – Medium: as many as 200 enclosures – Hyperscale – millions of servers

Current State • Power Usage Effectiveness PUE = Total facility power/IT equipment power • Data Center infrastructure Efficiency metric DCi. E 1/PUE • When first introduced, the US avg PUE was >2. 4 – For every Watt going to a server, 1. 4 Watts used in facility or by AC • Today, datacenters Google, Microsoft, etc. have PUE <1. 2. Some even less today ~ 1. 0 (http: //en. wikipedia. org/wiki/Power_usage_effectiveness)

Current State • Data center used to just keep up with demand, not keep energy bill low • Now, data center want to make as efficient and productive as possible • BUT – huge center of existing software still runs in highly inefficient data centers • Hyperscale cloud data centers (Google, etc. ) represent < 5% of US data centers • Small, medium, corporate, multitenant data centers still squander huge amounts of energy

Put everything into the cloud? • Lowest hanging fruit – efficiency of really small data centers (together millions of servers) – Make more efficient or – Replace with hyperscale facilities • Still a limit on efficiency • Still a new field of discovery

Current state (2014)- The Negative • Key findings from the NRDC-Anthesis report include: – Up to 30 percent of servers are “comatose” and no longer needed because projects have ended or business processes changed, but are still plugged in and consuming electricity. – Much of the energy consumed by U. S. data centers powers servers operating at 12 to 18 percent of capacity. Even sitting virtually idle, servers use significant amounts of power 24/7. – In 80 percent of organizations, the department responsible for data center management is separate from the one paying the electric bills. This “split incentive” reduces the likelihood of implementation of commonsense efficiency measures.

Current State – The Bad • The average desktop PC wastes half of the energy it consumes. • typical PC is in use only 4 hours each workday and sits idle for more than 5 hours. (http: //www. lbl. gov) • Computers and monitors represent 66% of the electrical load in the average office. • Most new PCs are ENERGY STAR rated but only 5% of PCs have features activated. (Energy Star) • PCs use about 10% more electricity than data centers. (California Energy Commission)

Current approaches • Replace old computers with new more energyefficient • But manufacturing through day-to-day uses energy • Dell , IBM, Cicso, Adobe, Apple listed as Green • What is “Greenest computer ever” ? – Is Mac. Book air (pro) greenest?

Goals for Future 1. Consider energy to manufacture, operate, dispose of 2. Sense and optimize world around us 3. Predict and respond to future events by modeling behavior (grown in performance) 4. Benefit of digital alternative to physical activities – E-newspapers, online shopping • Personal energy meter? ?