Methods for feedback-based optimal workload scheduling and devices thereof

A workload scheduler device, comprising memory comprising programmed instructions stored thereon and one or more processors configured to be capable of executing the stored programmed instructions to:

• determine one of a plurality of task categories for a received task;
• obtain a stored expected runtime for each of a plurality of CPUs to execute one standard computation unit (SCU) in the determined one of the plurality of task categories;
• select one of the plurality of CPUs based on the stored expected runtime;
• and
• dispatch the task to the selected one of the plurality of CPUs for execution,
• wherein the expected runtime for each of the plurality of CPUs is stored in a matrix and the one or more processors are further configured to be capable of executing the stored programmed instructions to insert a default expected runtime in a plurality of entries of the matrix in order to initialize the matrix, each of the entries of the matrix corresponding to another one of the plurality of task categories and another one of the plurality of CPUs, and
• wherein the plurality of task categories includes at least one from among a first task category that includes CPU-intensive tasks from batch workloads, a second task category that includes input/output-intensive tasks from interactive workloads, a third task category that includes blended tasks, and a fourth task category that includes Monte Carlo simulation tasks.

A non-transitory machine readable medium having stored thereon instructions for improved task scheduling across central processing units (CPUs) comprising executable code which when executed by one or more processors, causes the one or more processors to:

• determine one of a plurality of task categories for a received task;
• obtain a stored expected runtime for each of a plurality of CPUs to execute one standard computation unit (SCU) in the determined one of the plurality of task categories;
• select one of the plurality of CPUs based on the stored expected runtime;
• and
• dispatch the task to the selected one of the plurality of CPUs for execution,
• wherein the expected runtime for each of the plurality of CPUs is stored in a matrix and the executable code when executed by the processors further causes the one or more processors to insert a default expected runtime in a plurality of entries of the matrix in order to initialize the matrix, each of the entries of the matrix corresponding to another one of the plurality of task categories and another one of the plurality of CPUs, and
• wherein the plurality of task categories includes at least one from among a first task category that includes CPU-intensive tasks from batch workloads, a second task category that includes input/output-intensive tasks from interactive workloads, a third task category that includes blended tasks, and a fourth task category that includes Monte Carlo simulation tasks.