Steering compute workloads based on carbon costs

A system, comprising:

• one or more processors and one or more non-transitory memories storing computer-readable instructions that, when executed by the one or more processors, cause the one or more processors to:
• collect, by a management interface configured to monitor energy consumption of servers in a cloud computing environment, energy consumption information associated with a plurality of servers processing workloads in the cloud computing environment by obtaining, by the management interface, instantaneous energy consumption values for central processing units (CPUs), graphics processing units (GPUs), and memory devices of the servers;
• calculate, by an energy consumption manager in communication with the management interface, carbon cost metrics for the plurality of servers based on the energy consumption information, the carbon cost metrics being weighted according to different types of servers;
• identify, by the energy consumption manager, a subset of the plurality of servers as being underutilized based on the carbon cost metrics and hardware utilization ratios;
• calculate, by the energy consumption manager, an overall carbon cost of the plurality of servers based on the carbon cost metrics;
• generate, by a workload manager, container platform operation labels for the plurality of servers responsive to the overall carbon cost being below the threshold value, the container platform operation labels comprising labels indicating whether a server is underutilized or available;
• label the identified subset of the plurality of servers with container platform operation labels indicating that the identified subset of the plurality of servers is underutilized;
• label one or more of the plurality of servers not identified as being underutilized with container platform operation labels indicating that the one or more of the plurality of servers is available;
• and
• reducing energy consumption and carbon emissions of the cloud computing environment by migrating, by a non-transitory computer-readable medium on one or more of the servers based on the container platform operation labels, a workload executing on a server labelled as underutilized to a server labelled as available.

A non-transitory computer-readable medium comprising computer-executable instructions stored thereon that, when executed by at least one processor, cause the at least one processor to:

• collect, by a management interface configured to monitor energy consumption of servers in a cloud computing environment, energy consumption information associated with a plurality of servers processing workloads in the cloud computing environment by obtaining, by the management interface, instantaneous energy consumption values for central processing units (CPUs), graphics processing units (GPUs), and memory devices of the servers;
• calculate, by an energy consumption manager in communication with the management interface, carbon cost metrics for the plurality of servers based on the energy consumption information, the carbon cost metrics being weighted according to different types of servers;
• identify, by the energy consumption manager, a subset of the plurality of servers as being underutilized based on the carbon cost metrics and hardware utilization ratios;
• calculate, by the energy consumption manager, an overall carbon cost of the plurality of servers based on the carbon cost metrics;
• generate, by a workload manager, container platform operation labels for the plurality of servers responsive to the overall carbon cost being below the threshold value, the container platform operation labels comprising labels indicating whether a server is underutilized or available;
• label the identified subset of the plurality of servers with container platform operation labels indicating that the identified subset of the plurality of servers is underutilized;
• label one or more of the plurality of servers not identified as being underutilized with container platform operation labels indicating that the one or more of the plurality of servers is available;
• and
• reducing energy consumption and carbon emissions of the cloud computing environment by migrating, by a non-transitory computer-readable medium on one or more of the servers based on the container platform operation labels, a workload executing on a server labelled as underutilized to a server labelled as available.