Data Center Technology: Physical Infrastructure IT Trends Affecting New Technologies and Energy Efficiency Imperatives in the Data Center Hisham Elzahhar Regional Enterprise & System Manager, Schneider Electric IT business EMEA, Dubai
Keystrokes Kilowatts Heat OUT Electricity IN 2
US Electrical Energy Sources 2006 Other Renew ables Hydro-Electric 2% 7% Nuclear 19% Natural Gas 20% Petroleum 2% CoalCoal 50% Coal Petroleum Natural Gas Nuclear Hydro-Electric Other Renewables Source US EIA 3
Prime Electrical Source 4
WHICH infrastructure? BUILDING infrastructure Building systems HVAC Electrical system Fire suppression Lighting Security BMS DATA CENTER infrastructure Power Cooling Racks Management Lighting Fire suppression Physical security IT infrastructure IT assets Servers, storage hypervisors,, NMS NETWORK infrastructure Switches, cabling, routers 5
WHICH infrastructure? Focus of this discussion BUILDING infrastructure Building systems HVAC Electrical system Fire suppression Lighting Security BMS DATA CENTER infrastructure Power Cooling Racks Management Lighting Fire suppression Physical security IT infrastructure IT assets Servers, storage hypervisors,, NMS NETWORK infrastructure Switches, cabling, routers 6
Data center planning and operation is under increasing pressures Increasing availability Rapid changes in expectations IT technology Uncertain long-term plans for Energy and service capacity or density cost control pressure High density blade server power/heat Dynamic power variation Regulatory requirements Server consolidation In response, will need to change the way the world designs, installs, operates, manages, and maintains data centers 7
The increasing power density of data centers Management challenge: HIGH DENSITY Power density of IT devices is leveling off but power density of data centers data centers continues to increase due to packing of high-density devices into smaller floor footprint 2000 2009 KW per rack continues to increase, raising the need for management to keep things under control 8
High density is stressing Management challenge: HIGH DENSITY power and cooling systems IT is getting boxed-in by limitations of power and cooling infrastructure High density increases the risk of unpredictable cooling Capacity is tight in some places, unused and unusable ( stranded ) in others High density requires informed and efficient allocation of your expensive power/cooling resources High density increases the need to know where new devices can be squeezed in to available capacities 9
The Newest Challenge: EFFICIENCY Efficiency goal: Provide power and cooling in the amount needed, when needed, and where needed but no more than what is required for redundancy and safety margins But we can t manage what we can t measure 10
Datacenter Efficiency - DCiE Data center Power to data center Power path to IT POWER system Power to IT IT equipment Power to Secondary Support COOLING system Physical infrastructure* *To simplify the analysis, subsystems consuming a small amount of power are not included in this discussion: Cabling Physical security Switches Generator Lights Switchgear White paper 113 = Data Center infrastructure Efficiency Power to IT Power to data center ( )% 11
Datacenter Efficiency Data Center Physical Infrastructure IT COOLING COOLING system system POWER POWER system system 12
Power Chain Losses 4,930 barrels 47 tons SO2 16 tons N2O 6,539 tons CO2 Per mw/yr 1mW DCiE @ 47% 45 racks @ 10kW 13
Inefficiencies Create Consumption Computing inefficiencies > More servers Server inefficiencies > More power and cooling Power and cooling inefficiencies > More power consumption Inefficiencies drive both power consumption and material consumption 14
Primary drivers of inefficiency Oversizing of power and cooling equipment Pushing cooling systems to cool densities higher than they were designed for Ineffective room layout Ineffective airflow patterns Redundancy (for availability) Inefficient power and cooling equipment Inefficient operating settings of cooling equipment Clogged air or water filters Disabled or malfunctioning cooling economizer modes Raised floor clogged with wires 15
Efficiency: key reference points More than 50% of the power going into a typical data center goes to the power and cooling systems NOT to the IT loads The typical 1MW (IT load) data center is continuously wasting about 400kW or 2,000 tons of coal per year due to poor design (DCiE = 50%, instead of best-practice 70%) Every kw saved in a data center saves about $1,000 per year Every kw saved in a data center reduces carbon dioxide emissions by 5 tons per year Every kw saved in a data center has a carbon reduction equivalent to eliminating about 1 car from the road. A 1% improvement in data center infrastructure efficiency (DCiE) corresponds to approximately 2% reduction in electrical bills References: APC White Paper 66 16
Power tools for The Efficient Enterprise 17
Power tools The Four Cs 1 2 omponents MODULAR and SCALABLE, with best-in-class EFFICIENCY lose-coupled cooling Placement of cooling units near the heat source 3 ontainment Thermal containment of airflow in high-density zones 4 apacity management Instrumented intelligence to optimize use of power and cooling capacity 18
1 omponents with the right stuff Best-in-class component EFFICIENCY Efficient Agile MODULAR SCALABLE component design Scalable External modularity Internal modularity 19
Problem: Underloading Low loading = low efficiency In a traditional data center, over half the power consumption of the power/cooling infrastructure is fixed and does not go down when IT load goes down Efficiency degrades as IT load declines Underloading is a primary contributor to inefficiency Data center Efficiency E ffic ie n c y 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Efficiency degrades at low loads Typical load range 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% % IT Load IT load 20
Solution: Right-sizing Efficiency gain through modular scalable buildout avoids oversizing / underloading Data center Efficiency Efficiency 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Power and cooling installation method 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% % IT load Load 21
Modular scalable design Reduce power consumption up to 30% by right right-sizing sizing power and cooling infrastructure Avoid underloading run more efficiently Pay only for what you need, when you need it P = Power C = Cooling R = Racks 22
500kW of scalable, high-efficiency power protection 100kW 125kW 150kW 175kW 200kW 225kW 250kW 275kW 300kW 325kW 350kW 375kW 400kW 425kW 450kW 500kW 475kW 23
lose-coupled cooling Reduce power consumption up to 20% with InRow architecture Closely couples cooling with heat load, preventing exhaust air recirculation Less fan power than traditional raised-floor system Varying equipment temperatures are constantly held to set point conditions Lowers operating cost by monitoring inlet temperatures to modulate cooling capacity based on the cooling demand Fan speed adjusts to follow changing IT heat load 24
Close-coupled cooling InRow air conditioner Heat captured and rejected to chilled water Cold air is supplied to the cold aisle Hot-aisle air enters from rear, preventing mixing Hot aisle Cold aisle Can operate on hard floor or raised floor 25
Efficiency comparison 100% 90% Cooling Efficiency Cooling efficiency 80% 70% 60% 50% 40% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% IT load % IT Load Cooling efficiency = useful cooling power / (power consumed + useful cooling power) 26
ontainment Eliminate expensive temperature cross-contamination contamination with thermal containment options Simplifies analysis and understanding of the thermal environment Hot Aisle Containment (HAC) Increases predictability of the cooling system Increases cooling EFFICIENCY and cooling CAPACITY by returning warmest possible air to cooling units Ensures proper air distribution by separating supply and return air paths Rack Air Containment (RAC) 27
Rack Air Containment Rear Rear Containment Rear containment prevents hot exhaust air from escaping InRow cooling unit InRow cooling unit All exhaust air is returned to InRow cooling unit Optional front containment directs cool air to front of servers Allows up to 60 kw per rack (30 kw with N+1 redundancy) NetShelter SX rack Front Containment Front Top Down View 28
Hot aisle containment vs traditional room cooling Inherently higher power density capability than room designs Fan power is reduced by 50% Needless dehumidification / re-humidification is eliminated Need for high-bay areas and raised floors is reduced or eliminated (particularly for small installations) Cooling capacity can follow IT loads that move due to virtualization and server power management C o o l i n g E f f i c i e n c y 100% 90% 80% 70% 60% 50% 40% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% % IT Load Cooling efficiency = useful cooling power / (power consumed + useful cooling power) 29
Hot Aisle Containment areas can be added as needed 30
apacity Management Increase IT staff efficiency with predictable Capacity Management Identify over- and under-utilized areas of your data center Minimize waste and human error via predictable software monitoring, sensing, and environmental control Quickly adapt to change with real-time data on what to power and where to cool 31
Capacity Manager Physical equipment provisioning Quickly locate the optimum spot for that next server based on space, cooling, and power needs Rack elevations Easy-to-use front view for accurate and detailed representation of equipment layout Airflow analysis Locate new devices without overheating new or existing equipment by simulating changes in; supply temperature, airflow and number of cooling units Available capacity Understand available capacity by calculating actual space, power and cooling consumption against data center architecture constraints Capacity grouping Specify architecture capabilities to; match IT equipment with availability needs ad avoid stranded space, power and cooling capacity Design analysis Model the effects of and compare alternative layouts through detailed design analysis 32
Capacity and energy management Poor utilization of capacity is a primary cause of inefficiency Software can identify available capacity (even by rack) and help prevent creation of stranded capacity Side effect is you can fit more IT equipment in the power and cooling envelope of the data center Energy management can identify efficiency improvement opportunities Infrastructure Central Software With Capacity Manager 33
Power consumptions compared to the IT load IT Load Aux Devices Lights Humidifier Chiller Pumps Heat Rejection CRAC Distribution Wiring Switchgear Generator Improving efficiency means working to reduce power consumption (increase efficiency) for each of these device categories PDU UPS Reference: APC White Paper 114 0% 20% 40% 60% 80% 100% 120% Power Power consumption Consumption as as % of % the of IT IT Load load 0.0% 20.0% 40.0% 60.0% 80.0% 100.0% 120.0% Data for a typical tier 4 data center operating at 30% of rated load 34
Drivers of infrastructure efficiency gains (Baseline: Average of existing installed base) IMPROVEMENT Device Gain DCiE Gain $$ saved over 15 years in a 1MW data center** Move from room cooling to dynamic row/rack cooling 70% 8% $5,900,000 Cooling economizers 38% 4% $2,500,000 Right-sizing through modular power and cooling equipment 4% 4% $2,400,000 Higher UPS efficiency 8% 4% $1,900,000 415/240 V transformerless power distribution (NAM)* Dynamic control of cooling plant (VFD fans, pumps, chillers) 4% 2.5% $1,500,000 25% 2.5% $1,200,000 TOTAL to get industry from 47% to 72% DCiE 25% 25% $14,700,000 *No benefit outside of NAM; Transformer based PDUs typically in NAM only **$$ values based on $.15 per kwh electric cost, starting DCiE of 47%, ave density 8KW/rack 35
Power Chain Losses Could Be 4,930 barrels 6,539 tons CO2 47 tons SO2 16 tons N2O Per mw/yr 1,971 barrels 2,615 tons CO2 19 tons SO2 6 tons N2O 1mW 400kW DCiE @ 70% 36
Visit us Hall 4, booth 4405 (next to the BP Carbon Theater) Tour the booth and be entered into our daily lottery The lucky winner to receive a brand new Amazon Kindle e-reading device right away! 37
Questions? 38