Workload prediction for network-based computing

摘要

Technologies are generally described for systems and methods effective to efficiently schedule a workload in a cloud computing system. In one example, calendar data is collected from respective sets of devices associated with respective sets of subscribers and a workload to be performed at a specific time or range of time is predicted based in part on an analysis of calendar data. Moreover, timing data associated with a set of predicted requests is determined based on the analysis and at least a portion of cloud computing resources that service the set of predicted requests are dynamically scheduled based on timing data.

主权项

1. A method, comprising:
receiving, by a system comprising a processor, first calendar data transmitted from respective first sets of devices, wherein the first calendar data comprises respective first meeting data associated with first sets of subscribers of the first sets of devices; prior to receiving a first set of requests from the first sets of devices, determining a workload to be performed at a specific time or range of time based in part on an analysis of the first calendar data; and based on the workload and start-up delay data associated with servers, allocating a first set of the servers to respond to the first set of requests and a second set of the servers to respond to a second set of requests received from respective second sets of devices that do not share second calendar data with the system, wherein the second calendar data comprises respective second meeting data associated with a second set of subscribers of the second set of devices and wherein the first set of the servers have a first start-up delay that is greater than a second start-up delay of the second set of the servers.