Cooperative batch scheduling in multitenancy system based on estimated execution time and generating a load distribution chart

摘要

Systems and methods for cooperative batch scheduling in multitenancy computing systems are described. A number of batch requests are received in the computing system, where each batch request includes a job reference, and a start time when the execution of the referenced job to be launched. Duration of execution is estimated for each job referenced by the requests. The estimation is based on predefined criteria that include a product of median execution times for at least one recurring operation. An anonymous load chart is created based on the start times and the estimated duration of execution of each job referenced by the plurality of batch requests. The anonymous load chart may take into account the available capacity of the computing system. The anonymous load chart is exposed to a number of isolated users of the computing system for cooperative batch scheduling.

主权项

1. A non-transitory computer readable storage medium storing instructions, which when executed by a computer, cause the computer to:
receive a plurality of job requests at a shared computer system from one or more of a plurality of isolated tenants of the shared computer system, wherein each job request of the plurality of job requests includes a reference to a corresponding job of a plurality of jobs and a start time for beginning an execution of the referenced job; for each job of the plurality of jobs referenced by the plurality of job requests, estimate, by the shared computer system, a duration of execution of the job based on a type of the job, wherein estimating the duration of execution of a job comprises:
identifying a plurality of objects of a specific type to be processed during the execution of the job,assigning a median duration for processing an object of the specific type, wherein assigning the median duration comprises:
estimating the median duration in accordance with an expected capacity of the shared computer system during the execution of the job, andadding, to the duration of execution of the job, a result of a multiplication between the median duration and a cardinality of the plurality of objects; create a distribution over a future period of time of load levels of the shared computer system resulting from the start times and the estimated durations of execution of the plurality of jobs referenced by the plurality of job requests; and expose a load chart of the distribution of the load levels of the shared computer system to at least one of the plurality of isolated tenants of the shared computer system to enable collaborative job scheduling between the isolated tenants for the future period of time.