Assigning data processing workloads in data centers based on non-data processing overhead

A computer program product comprising:

• a non-transitory computer readable storage medium comprising computer readable program code embodied in the medium that when executed by at least one processor of a computer system causes the computer system to perform operations comprising:
• for each respective rack of a plurality of racks each comprising a plurality of servers:
• predicting power consumption of the respective rack based on data processing demands that are placed on the plurality of servers for a given data processing workload by:
• determining a normalized respective power utilization index defining an amount of energy that is consumed by a respective one of the plurality of servers in the respective rack for each of the respective ones of the plurality of servers in the respective rack for a unit of workload performed by the respective one of the plurality of servers in the respective rack by:
• obtaining, via a user interface of the data processing apparatus connected via a network to the respective one of the plurality of servers, measurements of power consumed by the respective one of the plurality of servers in response to a plurality of workloads and measurements of workload demand placed on the respective one of the plurality of servers for the plurality of workloads;
• storing the measurements of power and measurements of workload demand in a memory of the data processing apparatus;
• and
• determining the normalized respective power utilization index for the respective one of the plurality of servers based on regression analysis of the measurements of power consumed and/or the measurements of workload demand and storing the normalized respective power utilization index in the memory;
• predicting respective power consumption of the respective ones of the plurality of servers in the respective rack for the future data processing workload, based on the normalized respective power utilization indexes for the plurality of servers in the respective rack;
• and
• predicting power consumption of the respective rack based upon the respective power consumption of the plurality of servers in the respective rack;
• measuring, by the data processing apparatus, power consumed by the respective rack when the plurality of servers is performing the given data processing workload;
• deriving, by the data processing apparatus, a metric of power consumed by the respective rack for non-data processing overhead based on a difference between results of the predicting and the measuring;
• determining, by the data processing apparatus, an availability for processing a future workload by the respective rack based on the metric of power consumed by the respective rack for the non-data processing overhead;
• and
• configuring the data processing apparatus, based on the availability, to:
• identify a plurality of racks among the plurality of racks that have the availability;
• select a rack of the plurality of racks that have the availability and based on the metric of power consumed by the rack for the non-data processing overhead and a time period that reduces a cost of power consumed by the rack;
• select a server in the rack from among the plurality of servers of the rack that reduces the overall power consumed for the non-data processing overhead in processing the future workload;
• transmit a command to the rack to process the future data processing workload by the selected server in the rack;
• and
• receive an indication that the rack is processing the workload at the time period that reduces the cost of power consumed by the rack.

A computer system comprising:

• a processor;
• and
• a workload assigning system that runs on a processor, the workload assigning system configured to perform operations to:
• for each respective rack of a plurality of racks each comprising a plurality of servers:
• predict power consumption of the respective rack based on data processing demands that are placed on the plurality of servers of the respective rack for a given data processing workload by:
• determining a normalized respective power utilization index defining an amount of energy that is consumed by a respective one of the plurality of servers in the respective rack for each of the respective ones of the plurality of servers in the respective rack for a unit of workload performed by the respective one of the plurality of servers in the respective rack by:
• obtaining, via a user interface of the data processing apparatus connected via a network to the respective one of the plurality of servers, measurements of power consumed by the respective one of the plurality of servers in response to a plurality of workloads and measurements of workload demand placed on the respective one of the plurality of servers for the plurality of workloads;
• storing the measurements of power and measurements of workload demand in a memory of the data processing apparatus;
• and
• determining the normalized respective power utilization index for the respective one of the plurality of servers based on regression analysis of the measurements of power consumed and/or the measurements of workload demand and storing the normalized respective power utilization index in the memory;
• predicting respective power consumption of the respective ones of the plurality of servers in the respective rack for the future data processing workload, based on the normalized respective power utilization indexes for the plurality of servers in the respective rack;
• and
• predicting power consumption of the respective rack based upon the respective power consumption of the plurality of servers in the respective rack;
• measure power consumed by the respective rack when the plurality of servers is performing the given data processing workload;
• derive a metric of power consumed by the respective rack for non-data processing overhead based on a difference between results of the predicting and the measuring;
• determine an availability for processing a future workload by the respective rack based on the metric of power consumed by the respective rack for the non-data processing overhead;
• and
• identify a plurality of racks among the plurality of racks that have the availability;
• select a rack of the plurality of racks that have the availability and based on the metric of power consumed by the rack for the non-data processing overhead and a time period that reduces a cost of power consumed by the rack;
• select a server in the rack from among the plurality of servers that reduces the overall power consumed for the non-data processing overhead in processing the future workload;
• transmit a command to the rack to process the future data processing workload by the selected server in the rack;
• and
• receive an indication that the rack is processing the workload at the time period that reduces the cost of power consumed by the rack.