量化连续体：混合计算的成本、价值和策略.pdf

1、Quantifying the Quantifying the Continuum:Continuum:Scott GodwinDirector,Center for Continuum ComputingDr.Roger CassChief Architect,Continuum ComputingCost,Value and Strategyfor Hybrid ComputingThen“What systemshould webuild next?”Now“What computing environment best enables the science?”Rethinking H

2、ow We Invest in Scientific ComputingNOWTodays Scientific Workflows Dont Livein One PlaceSimulationAIData-IntensiveWorkflowsEdgeInstrumentsCloud-NativeToolingHPCBut the cost models,value models,and planning assumptions often still reflect a world where compute is mostly on-prem and homogeneous.Goal:B

3、uild a framework to evaluate cost,value,and strategic fit across the computing continuumScience Has Outgrown a Single EnvironmentCloudOn-Prem HPCSpecialized&Emerging PlatformsEvery point on this continuum involves tradeoffs:PerformanceCostFlexibilityAgilityThroughputControlOptimizationInnovationEdge

4、The hardest HPC question isnt“Can we run it?”Its“Whereshould it run,and why?”NOWMost Existing Decision Models Stop Too EarlyWhat is OftenMeasuredWhat is OftenUnder-measuredWhat is OftenMissing Entirely Capital cost Hardware acquisition Utilization Run cost Foundational infrastructure costs Staff tim

5、e/Operational burden Queue delay/Access latency Workflow friction Data movement/integration cost Technology currency Opportunity cost Time-to-insight Scientific value Research optionality Ability to scale collaboration Readiness for AI and hybrid workflowsNOW“What computing environment best enables

6、the science?”The Framework We Set Out to Build“What does it really cost?”“What does it really cost?”ComputingCost Model“What scientific value does it create?”“How do institutions evolve towards this future?”ScientificValue ModelContinnum Strategy FrameworkComp