DATA AND TASK REALLOCATION IN DISTRIBUTED COMPUTING PLATFORMS

摘要

Technical solutions are described for optimizing operation of a server cluster. An example method includes receiving a job request that executes using a set of data blocks, the job request being associated with an expected completion time. The cluster server is used to identify a set of replica servers, wherein each server from the set of replica servers contains the set of data blocks. In response to each server from the set of replica servers estimating a completion time for the job request that is more than the expected completion time, a new server is initiated, the set of data blocks is relocated from a first server from the set of replica servers to the new server, and the job request is allocated to the new server.