Compute load shaping using virtual capacity and preferential location real time scheduling

Patent metadata

Field	Value
Publication number	US-11221595-B2
Application number	US-201916684013-A
Country	US
Kind code	B2
Filing date	Nov 14, 2019
Priority date	Nov 14, 2019
Publication date	Jan 11, 2022
Grant date	Jan 11, 2022

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for shaping compute load using virtual capacity. In one aspect, a method includes obtaining a load forecast that indicates forecasted future compute load for a cell, obtaining a power model that models a relationship between power usage and computational usage for the cell, obtaining a carbon intensity forecast that indicates a forecast of carbon intensity for a geographic area where the cell is located, determining a virtual capacity for the cell based on the load forecast, the power model, and the carbon intensity forecast, and providing the virtual capacity for the cell to the cell.

First claim

What is claimed is:

1. A computer-implemented method comprising:
obtaining a load forecast for a future period of time that indicates forecasted future compute load for a cell during respective intervals of the future period of time;
obtaining a power model that models a relationship between power usage and computational usage for the cell;
obtaining a carbon intensity forecast that indicates a forecast of carbon intensity for a geographic area where the cell is located for the respective intervals of the future period of time, wherein the carbon intensity forecast indicates that the carbon intensity for the geographic area where the cell is located is greater during a first interval of the respective intervals than a second interval of the respective intervals;
determining a virtual capacity for the cell for the respective intervals of the future period of time based on the load forecast, the power model, and the carbon intensity forecast,
wherein determining the virtual capacity for the cell includes determining to shift a portion of the forecasted future compute load from the first interval of the respective intervals to the second interval of the respective intervals based on that the carbon intensity forecast indicates that the carbon intensity for the geographic area where the cell is located is greater during the first interval of the respective intervals than the second interval of the respective intervals; and
providing the virtual capacity for the cell for the future period of time to the cell,
wherein, based on the virtual capacity, the cell determines to defer executing a job during the first interval and determines to execute the job during the second interval.

2. The method of claim 1 , wherein determining a virtual capacity for the cell for the respective intervals of the future period of time based on the load forecast, the power model, and the carbon intensity forecast comprises:
obtaining respective load forecasts that indicate forecasted future compute loads for respective multiple other cells;
obtaining respective power models that model relationships between power usage and computational usage for the respective other cells;
obtaining respective carbon intensity forecasts that indicate forecasts of carbon intensity for geographic areas where the respective other cells are located; and
determining both the virtual capacity for the cell and respective virtual capacities for the other cells based on the load forecast, the power model, the carbon intensity forecast, the respective load forecasts, the respective power models, and the respective carbon intensity forecasts.

3. The method of claim 2 , comprising:
receiving an indication of a total amount of spatially and temporally flexible demand,
wherein determining both the virtual capacity for the cell and respective virtual capacities for the other cells is further based on the total amount of spatially and temporally flexible demand.

4. The method of claim 2 , comprising
receiving an indication of a total amount of spatially and temporally flexible demand,
wherein a preference for deciding in which cell a job runs is determined to aid real-time load scheduling.

5. The method of claim 1 , wherein the virtual capacity indicates a maximum computational capacity for the cell for pre-determined time intervals.

6. The method of claim 1 , wherein the virtual capacity indicates a maximum computational capacity for the cell for each hour in a day.

7. The method of claim 1 , wherein the cell is configured to perform operations of:
receiving the job;
determining an amount of computational resources needed to execute the job;
determining that a difference between the virtual capacity for the cell and a current computational usage of the cell satisfies the amount of computational resources needed to execute the job; and
executing the job.

8. The method of claim 1 , wherein the cell is configured to perform operations of:
receiving the job;
determining an amount of computational resources needed to execute the job;
determining that a difference between the virtual capacity for the cell and a current computational usage of the cell does not satisfy the amount of computational resources needed to execute the job; and
deferring execution of the job.

9. The method of claim 1 , wherein determining a virtual capacity for the cell for the respective intervals of the future period of time based on the load forecast, the power model, and the carbon intensity forecast comprises:
determining, based on the load forecast, the power model, and the carbon intensity forecast, the virtual capacity such that the virtual capacity reduces load peaks in the cell and reduces carbon footprint from power usage by the cell.

10. The method of claim 1 , wherein obtaining a power model that models a relationship between power usage and computational usage for the cell comprises:
training the model based on historical power usage of the cell and historical computational capacity of the cell that is used.

11. The method of claim 1 , wherein obtaining a load forecast for a future period of time that indicates forecasted future compute load for a cell during respective intervals of the future period of time comprises:
determining a portion of the load forecast that is not time flexible, wherein determining the virtual capacity for the cell includes determining the virtual capacity to be greater than the portion of the load forecast that is not time flexible.

12. A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
obtaining a load forecast for a future period of time that indicates forecasted future compute load for a cell during respective intervals of the future period of time;
obtaining a power model that models a relationship between power usage and computational usage for the cell;
obtaining a carbon intensity forecast that indicates a forecast of carbon intensity for a geographic area where the cell is located for the respective intervals of the future period of time, wherein the carbon intensity forecast indicates that the carbon intensity for the geographic area where the cell is located is greater during a first interval of the respective intervals than a second interval of the respective intervals;
determining a virtual capacity for the cell for the respective intervals of the future period of time based on the load forecast, the power model, and the carbon intensity forecast,
wherein determining the virtual capacity for the cell includes determining to shift a portion of the forecasted future compute load from the first interval of the respective intervals to the second interval of the respective intervals based on that the carbon intensity forecast indicates that the carbon intensity for the geographic area where the cell is located is greater during the first interval of the respective intervals than the second interval of the respective intervals; and
providing the virtual capacity for the cell for the future period of time to the cell,
wherein, based on the virtual capacity, the cell determines to defer executing a job during the first interval and determines to execute the job during the second interval.

13. The system of claim 12 , wherein determining a virtual capacity for the cell for the respective intervals of the future period of time based on the load forecast, the power model, and the carbon intensity forecast comprises:
obtaining respective load forecasts that indicate forecasted future comp

Assignees

• Google Llc

Inventors

• Radovanovic Ana
• Nobrega Duarte Alexandre
• Roy Binayak
• Xiao Diyue
• Cirne Filho Walfredo Da Costa
• Haridasan Maya
• Koningstein Ross
• Mullen Eric Alexander

Classifications

• H02J2103/30

  Simulating, planning, modelling, reliability check or computer assisted design [CAD] of electric power networks · CPC title

• H02J3/003

  Load forecast, e.g. methods or systems for forecasting future load demand · CPC title

• G06Q10/04 Primary

  Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202) · CPC title

• G05B13/0265

  the criterion being a learning criterion · CPC title

• G06Q10/06314 Primary

  Calendaring for a resource · CPC title