Granted Patent B2

US 12,585,494 · App. 17/722,007 · Granted Mar 24, 2026

Inventors: William Jeffery White (Plano, TX); Said Tabet (Austin, TX)

Assignee: Dell Products L.P.

G06F9/4881G06F9/5038

Abstract

Claims (65)

1 . A method for performing domain level scheduling in a distributed multi- tiered computing (DMC) environment, comprising:

receiving, by a local controller, a scheduling package from a global controller, wherein:

the scheduling package specifies a scheduling job, and

the global controller sent the scheduling package as network data traffic units over a series of network devices that operatively connect the local controller and the global controller;

normalizing, by the local controller associated with a DMC domain, task resource demand dimensions for each task associated with the scheduling job, wherein the resource demand dimensions comprise:

compute units, wherein the compute units comprise a quantity of processors and processor cores;

memory units;

accelerator units, wherein the accelerator units comprise a quantity of graphics processing units (GPUs), a quantity of GPU memory, a quantity of field programmable gate arrays (FPGAs), and a quantity of FPGA logic elements; and

network units, wherein the network units comprise a quantity of network interface cards and a total network bandwidth;

summing the task resource demand dimensions for each task to generate resource demand vectors;

classifying tasks based on priority;

sorting tasks based on associated resource demand vectors;

obtaining critical path, earliest start time, and latest start time associated with each task;

sorting tasks based on the critical path and the earliest start time; and

generating scheduling assignments based on the priority, capacity of devices in a final candidate list, resource demand vectors, earliest start time, and the critical path.

2 . The method of claim 1 , wherein the tasks, the priority, and the task resource demand dimensions are specified by a scheduling package associated with the scheduling job.

3 . The method of claim 1 , wherein the earliest start times specify the earliest relative time the tasks may be scheduled.

4 . The method of claim 1 , wherein the latest start times specify the latest relative times the tasks may be scheduled.

5 . The method of claim 1 , wherein the critical path specifies a portion of the tasks to schedule on a same device.

6 . The method of claim 1 , wherein the DMC domain comprises one selected from a group consisting of:

an edge domain;

a core domain; and

a cloud domain.

7 . The method of claim 6 , wherein:

the edge domain comprises an edge domain device set;

the core domain comprises a core domain device set; and

the cloud domain comprises a cloud domain device set.

8 . A non-transitory computer readable medium comprising computer readable program code, which when executed by a computer processor enables the computer processor to perform a method for managing a distributed multi-tiered computing (DMC) environment, comprising:

receiving, by a local controller, a scheduling package from a global controller, wherein:

the scheduling package specifies a scheduling job, and

the global controller sent the scheduling package as network data traffic units over a series of network devices that operatively connect the local controller and the global controller;

normalizing, by the local controller associated with a DMC domain, task resource demand dimensions for each task associated with the scheduling job, wherein the resource demand dimensions comprise:

compute units, wherein the compute units comprise a quantity of processors and processor cores;

memory units;

accelerator units, wherein the accelerator units comprise a quantity of graphics processing units (GPUs), a quantity of GPU memory, a quantity of field programmable gate arrays (FPGAs), and a quantity of FPGA logic elements; and

network units, wherein the network units comprise a quantity of network interface cards and a total network bandwidth;

summing the task resource demand dimensions for each task to generate resource demand vectors;

classifying tasks based on priority;

sorting tasks based on associated resource demand vectors;

obtaining critical path, earliest start time, and latest start time associated with each task;

sorting tasks based on the critical path and the earliest start time; and

generating scheduling assignments based on the priority, capacity of devices in a final candidate list, resource demand vectors, earliest start time, and the critical path.

9 . The non-transitory computer readable medium of claim 8 , wherein the tasks, the priority, and the task resource demand dimensions are specified by a scheduling package associated with the scheduling job.

10 . The non-transitory computer readable medium of claim 8 , wherein the earliest start times specify the earliest relative time the tasks may be scheduled.

11 . The non-transitory computer readable medium of claim 8 , wherein the latest start times specify the latest relative times the tasks may be scheduled.

12 . A system for managing a distributed multi-tiered computing (DMC) environment, the system comprising:

a DMC environment; and

a local controller associated with a DMC domain of the DMC environment, comprising a processor and memory, and configured to:

receive, by the local controller, a scheduling package from a global controller,

wherein:

the scheduling package specifies a scheduling job, and

the global controller sent the scheduling package as network data traffic units over a series of network devices that operatively connect the local controller and the global controller;

normalize task resource demand dimensions for each task associated with a scheduling job, wherein the resource demand dimensions comprise:

compute units, wherein the compute units comprise a quantity of processors and processor cores;

memory units;

accelerator units, wherein the accelerator units comprise a quantity of graphics processing units (GPUs), a quantity of GPU memory, a quantity of field programmable gate arrays (FPGAs), and a quantity of FPGA logic elements; and

network units, wherein the network units comprise a quantity of network interface cards and a total network bandwidth;

sum the task resource demand dimensions for each task to generate resource demand vectors;

classify tasks based on priority;

sort tasks based on associated resource demand vectors;

obtain critical path, earliest start time, and latest start time associated with each task;

sort tasks based on the critical path and the earliest start time; and

generate scheduling assignments based on the priority, capacity of devices in a final candidate list, resource demand vectors, earliest start time, and the critical path.

13 . The system of claim 12 , wherein the tasks, the priority, and the task resource demand dimensions are specified by a scheduling package associated with the scheduling job.

14 . The system of claim 12 , wherein the earliest start times specify the earliest relative time the tasks may be scheduled.

Assignments (1)

Continuity (1)

References Cited (61)