Stranded Capacity in Data Centers is Costing the Earth.pdf

1、VP Sales,CadenceHassan MoezziThe Scale of Stranded CapacityThere are 7 million data centers globally,with 105 gigawatts of built capacity consuming 594 terawatt hours of energy every year.This equates to 2.4%of the global energy drawOver 30 gigawatts of data center power capacity worldwide remains u

2、nused.This is not economically or ecologically sustainable30GigawattsGigawatts105Source Courtesy of Cato DigitalThe Scale of Stranded CapacityEnterprise data centers utilize around 56%of their design capacity over their lifetime.Source Courtesy The 451 Digital Infrastructure AllianceSupply Chain Def

3、ining a Data CenterThe Data Center LifecycleDesign/18 MonthsOperations/240 Months+Data center lifecycle is 20+yearsDesigned to a theoretical IT load Based on 5-year projectionsDefined 2-3 years before initial startup The IT equipment lifecycle is 3-5 years18 months for processors4-6 years for storag

4、e and network gearEVERY change creates divergenceCommissioningThe commissioning process is continuousThe Elephant in the Data CenterStranded CapacityHow and Why is Stranded Capacity Created?Stranded capacity is created:Overprovisioning for RASDesign vs implementation diverge resulting in increases F

5、ragmentation of space,power,and coolingOverprovisioning(by design)CapacityAvailability(downtime risk)Efficiency(PUE)Intended operational performanceOf the two,fragmentation is by far the larger cause of stranded capacityStranded capacity is unavoidableData center operationsOverprovisioning(by design

6、)Stranded capacity from fragmentationOverprovisioning(by design)Performance gapFragmentation(Tetris Effect)in Data CentersFacility designFragmentation of Cooling in OperationsData center operationsFacility designThe Impact of FragmentationStranded capacity from fragmentationOverprovisioning(by desig

7、n)Performance gapTodays operational planning tools and techniques cannot visualize/quantify fragmentationOperators must overprovision capacities to account for fragmentationCapacity overprovisioning has knock-on effectsHigh energy costsExcessive carbon emissionsUnsustainable data centersHow can we r

8、educe/mitigate stranded capacity?Introducing the concept of a Introducing the concept of a Data Centers Digital TwinData Centers Digital TwinA paradigm shift in the lifecycle A paradigm shift in the lifecycle management of data centersmanagement of data centersWhy a Data Center Digital TwinBusiness

9、requirements change the dynamics of the data centerThese requirements/changes are random and unpredictable,but they all affect data center performanceHence it is necessary to predict the effects of these changes before physical implementationScientifically quantifies data center performance and prov

10、ides visibility of stranded capacity/fragmentationData Center Digital Twins are Calibrated and LivePhysics-based,virtual replica of the physical data centerCalibrated to match configuration and behavior of physical counterpartDynamically synchronized using real-time monitoring data from physical cou

11、nterpartCapacityAvailabilityEfficiencyOperational changesOutputOutputCarbon emissionsCapex/Opex savingsOptimizationData for training ML/AI surrogatesData Center Digital Twins Quantify PerformanceData Center Digital Twin Provide Industry-Scale InteroperabilityProject:consolidate IT configuration,save

12、 co-lo capacity,maximize the performance outcome of IT load placement Four options were simulated in Digital Twin to optimize performance Option 1 performs best:Large Healthcare Enterprise Operational Planning ExampleOption 1Option 2Option 3Option 4Transfer 75 kW of IT load(88 sqft)from co-lo to mai

13、n data hallCapacity provisionsOption 1Option 2Option 3Option 4Airflow(CFM)8,84413,28010,78411,179AvailabilityPASSFAILPASSFAILAvailability failAvailability fail19%more efficient than Option 3100%Availability(saves 25 kW of cooling capacity)Eliminates 75 kW of co-lo capacity(saves$250k per year in co-

14、lo costs)Lets Collaborate on solving for the Elephant In the DatacenterHassan Moezzi VP Sales,COCPs Data Center Optimization Initiative:Bridging the Gap for Performance and SustainabilityThe data center industry faces a critical challenge-stranded capacity-consuming energy,increasing costs,and contr

15、ibuting to environmental impacts.To address this issue,we are excited to announce the Data Center Optimization Initiative,a comprehensive educational campaign to:Raise awareness of stranded capacity and its root causes Define a scientific quantification of stranded capacity and data center performance Foster collaboration between stakeholders across the industryBy working together,we can break down the silos that cause stranded capacity and risk in data center operations.