Cross-platform workload processing

摘要

According to one aspect of the present disclosure, a system and technique for workload processing includes a host having a processor unit and a memory. A scheduler is executable by the processor unit to: receive a request to process a workload; access historical processing data to determine execution statistics associated with previous processing requests; determine whether the data of the workload is available for processing; in response to determining that the data is available for processing, determine whether a process for the workload is available; in response to determining that the process is available, determine resource availability on a computing platform for processing the workload; determine whether excess capacity is available on the computing platform based on the resource availability and the execution statistics; and in response to determining that excess capacity exists on the computing platform, initiate processing of the workload on the computing platform.

主权项

1. A system for cross-platform workload processing, comprising:
a plurality of computing platforms; a host having a processor unit and a memory; and a scheduler executable by the processor unit to:
receive a request to process a workload, the workload including data to be processed by different workload processes at various processing stages of the workload;determine a type of the workload;determine types of workload processes needed at the various processing stages for the determined type of workload;access historical processing data to determine historical execution statistics associated with previous processing requests for the determined type of the workload;determine, by accessing a data input location, whether the data of the type of workload is available for processing;in response to determining that the data is available for processing, determine, by accessing a process input location, whether a type of workload process associated with the data is available;in response to determining that the type of workload process associated with the data is available, identify available resources on each of the plurality of computing platforms based on observation of current workload processing levels of the plurality of computing platforms;determine whether the identified available resources on at least one of the plurality of computing platforms are sufficient to process the workload based on the historical execution statistics for the type of the workload;access a set of rules to determine a priority of the workload relative to other workloads based on the type of workload and a predicted resource requirement of the workload based on the historical execution statistics for processing on the computing platform;in response to determining the priority and that the identified available resources on the at least one of the plurality of computing platforms are sufficient to process the type of workload based on the historical execution statistics, select the at least one of the plurality of computing platforms; andinitiate processing of the workload on the selected computing platform.