| 主权项 |
1. A computer-implemented method of facilitating determination of job start times in a multi-cluster environment, said method comprising:
computing, by a processor, a start time for a job to be executed across multiple clusters of the multi-cluster environment, the computing using multiple phases of processing including:
determining, during a first phase of the multiple phases, a time region in which a job can execute in the multi-cluster environment, said time region comprising a period of time in which resource availabilities of a plurality of clusters of the multi-cluster environment are monotonically increasing in that resources are released during that period of time and resources are not consumed during that period of time;choosing, during a second phase of the multiple phases, for the determined time region, one or more selected points within the time region at which resource availability information is to be obtained, the choosing using an n-ary search, and the choosing comprising:
obtaining a list of resource release times for the plurality of clusters, wherein a resource release time is a point in time on a cluster when it is known that a resource is released;determining a fanout factor for the n-ary search, the fanout factor based, at least in part, on the list of resource release times; anddetermining, based on the fanout factor, a start of the determined time region, and an end of the determined time region, the one or more selected points within the time region;obtaining for the plurality of clusters resource availability information at the chosen one or more selected points within the time region; andusing the resource availability information of the plurality of clusters to determine a start time for the job to be executed across multiple clusters of the multi-cluster environment. |
