How to Make Profit: Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters Timo Hönig, Christopher Eibel, Adam Wagenhäuser, Maximilian Wagner, and Wolfgang Schröder-Preikschat 8th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS 2018), Tempe, AZ, USA. June 12, 2018
Gode Wind Farm: 582 MW North Sea, Europe Photo: DONG Energy
Solana Generating Station: 280 MW Arizona, USA Photo: Abengoa Energy
Solana Generating Station: 280 MW Arizona, USA Photo: Abengoa Energy, Google
Arizona State University: 24 MW Arizona, USA Photo: Arizona State University
Building a Runtime System for Heterogeneous HPC Clusters ☼ Reenewable electricity sources have a great impact on the grid. Availability of renewables leads to fmuctuating electricity prices. How can we design and build an HPC runtime system to exploit dynamic electricity pricing? Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 5
Building a Runtime System for Heterogeneous HPC Clusters ☼ Reenewable electricity sources have a great impact on the grid. Availability of renewables leads to fmuctuating electricity prices. How can we design and build an HPC runtime system to exploit dynamic electricity pricing? Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 5
Building a Runtime System for Heterogeneous HPC Clusters ☼ Reenewable electricity sources have a great impact on the grid. Availability of renewables leads to fmuctuating electricity prices. How can we design and build an HPC runtime system to exploit dynamic electricity pricing? Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 5
Energy Mix (Germany, 2017) Others 1 % Renewables 38 % Natural Gas 9 % Non-renewables 62 % Nuclear Power 13 % Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 6 dynamic availability of renewables fmuctuation electricity prices ■ non-renewables: 339 . 93 TWh, renewables: 209 . 97 TWh
Energy Mix (Germany, 2017) Others 1 % Renewables 38 % Natural Gas 9 % Non-renewables 62 % Nuclear Power 13 % Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 6 ■ non-renewables: 339 . 93 TWh, renewables: 209 . 97 TWh ■ dynamic availability of renewables fmuctuation electricity prices
Fluctuation of Electricity Prices Others Price [EUR/MWh] 0 non-renewable electricity sources non-renewable plus renewable electricity sources electricity price [week 43/2017] Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 7 1 % Mo Renewables 38 % Natural Gas 9 % Non-Renewables Nuclear Power 13 % …but renewables are often unpredictable consequence: strong fmuctuation of the electricity price 100 80 60 60 Tu We Th Fr Sa Su 20 40 80 40 Time [day of week] Power [GW] -100 -80 -60 -40 -20 0 20 extreme imbalance of supply & demand negative prices ■ non-renewables can be planned …
Fluctuation of Electricity Prices 7 100 Price [EUR/MWh] 0 non-renewable electricity sources non-renewable plus renewable electricity sources electricity price [week 43/2017] Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters Others Mo 1 % Renewables Natural Gas 9 % Non-Renewables Nuclear Power 13 % consequence: strong fmuctuation of the electricity price 80 60 40 20 Tu We Th Fr Sa Su 20 40 60 80 Time [day of week] Power [GW] -100 -80 -60 -40 -20 0 extreme imbalance of supply & demand negative prices ■ non-renewables can be planned … ■ …but renewables are often unpredictable
Fluctuation of Electricity Prices 0 40 60 80 100 Price [EUR/MWh] negative price non-renewable electricity sources Mo non-renewable plus renewable electricity sources electricity price [week 43/2017] Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters 7 non-renewables can be planned … …but renewables are often unpredictable 20 0 -20 20 Tu We Th Fr Sa Su 40 -40 60 80 Time [day of week] Power [GW] -100 -80 -60 ■ consequence: strong fmuctuation of the electricity price ■ extreme imbalance of supply & demand negative prices
Building a System that Benefjts from Dynamic Electricity Prices 10:00 8 Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters Penalty zone II Under Demand / Excess Supply Target zone Penalty zone I Excess Demand / Under Supply Power Demand HPC Cluster Power Demand [kW] Time [hours] 8 6 4 2 0 16:00 15:00 14:00 13:00 12:00 11:00 ■ challenge: take advantage of low and negative prices ■ consider contracts (i.e., penalties enforced by grid operators) ■ integrate with existing HPC cluster infrastructure
Building a System that Benefjts from Dynamic Electricity Prices 10:00 reduce energy demand when prices are high electricity pricing into its operating decisions build a fmexible runtime system that integrates data on 8 Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters Penalty zone II Under Demand / Excess Supply Target zone Penalty zone I Excess Demand / Under Supply Power Demand HPC Cluster Power Demand [kW] Time [hours] 11:00 12:00 13:00 14:00 15:00 16:00 0 2 4 6 8 increase energy demand when prices are low (or negative) ■ challenge: take advantage of low and negative prices ■ consider contracts (i.e., penalties enforced by grid operators) ■ integrate with existing HPC cluster infrastructure
Building a System that Benefjts from Dynamic Electricity Prices 10:00 Our proposed runtime system Albatross: 8 Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters Penalty zone II Under Demand / Excess Supply Target zone Penalty zone I Excess Demand / Under Supply Without Albatross With Albatross Power Demand [kW] Time [hours] 8 6 4 2 0 16:00 15:00 14:00 13:00 12:00 11:00 ■ implements low- and high-power operation modes ■ considers heterogeneity aspects of the cluster hardware ■ respects QoS and non-functional requirements of workloads
Albatross: Design and Implementation . Compute Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters . . . Compute … Power meters . . Compute Master node Cluster manager Job scheduler Compute Compute . . . Compute … … manager Job 9 Cluster Master node constraints Electricity & price . . . . . . constraints QoS scheduler ■ foobar Resource group 1 Job 1 Exec n 1 Exec 1 Probe 1 node 1 Job 2 Exec 1 Exec n 2 Probe 2 node 2 Resource group l Job m Exec 1 Exec n k Probe k node k ■ generic system model of an HPC cluster without Albatross Execution unit 1 Job 1 node 1 Job 2 Execution unit 2 node 2 Job m Execution unit n node n
Albatross: Design and Implementation . Compute Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters . . . Compute … Power meters . . Compute Master node Cluster manager Job scheduler Compute Compute . . . Compute … … manager Job 9 Cluster Master node constraints Electricity & price . . . . . . constraints QoS scheduler ■ extended system model of an HPC cluster with Albatross Resource group 1 Job 1 Exec n 1 Exec 1 Probe 1 node 1 Job 2 Exec 1 Exec n 2 Probe 2 node 2 Resource group l Job m Exec 1 Exec n k Probe k node k ■ generic system model of an HPC cluster without Albatross Execution unit 1 Job 1 node 1 Job 2 Execution unit 2 node 2 Job m Execution unit n node n
Albatross: Design and Implementation Compute . . … . Compute Compute Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters . . . … scheduler Job manager Cluster governor . 9 meters Power QoS Resource constraints . . Electricity & price constraints Master node … ■ extended system model of an HPC cluster with Albatross Resource group 1 Job 1 Exec n 1 Exec 1 Probe 1 node 1 Job 2 Exec 1 Exec n 2 Probe 2 node 2 Resource group l Job m Exec 1 Exec n k Probe k node k ■ input interfaces for job submission and control ■ submission interface for electricity-pricing data ■ real-time data as provisioned by grid operator ■ detailed job description, added QoS constraints
Albatross: Design and Implementation Compute . . … . Compute Compute Exploiting Fluctuating Electricity Prices with Albatross, A Runtime System for Heterogeneous HPC Clusters . . . … scheduler Job manager Cluster governor . 9 meters Power QoS Resource constraints . . Electricity & price constraints Master node … ■ extended system model of an HPC cluster with Albatross Resource group 1 Job 1 Exec n 1 Exec 1 Probe 1 node 1 Job 2 Exec 1 Exec n 2 Probe 2 node 2 Resource group l Job m Exec 1 Exec n k Probe k node k ■ power monitoring and control interfaces ■ probes operate power-management features (e.g., power capping) ■ hardware-specifjc metering (e.g., RAPL, measuring devices) ■ delivery of base data for feedback loop to the resource governor
Recommend
More recommend
