Granted Patent B2

US 11,159,609 · App. 16/833,435 · Granted Oct 26, 2021

Inventors: Michael J. McGrath (Virtinia, IE); Daire Healy (Enfield, IE); Christopher D. Lucero (Chandler, AZ); Marcin Spoczynski (Lexilip, IE)

Assignee: Intel Corporation

H04L67/1008G06F9/5083G06F9/547H04L67/1023G16Y40/10G16Y40/35

Abstract

Claims (133)

1. A non-transitory computer-readable storage medium of an edge computing system to identify a target edge node for deployment of a workload thereon, the computer-readable storage medium comprising computer-readable instructions that, when executed, cause at least one processor to perform operations comprising:

determining whether respective ones of candidate target edge nodes of a set of candidate target edge nodes of the edge computing system support workload determinism key performance indicators (KPIs) of the workload;

in response to a determination that one or more candidate target edge nodes support the workload determinism KPIs, selecting a target edge node from the one or more candidate edge nodes;

causing the workload to be deployed at the target edge node; and

parsing one or more fields of a Virtualized Industrial Function Descriptor (VIFD) to determine the workload determinism KPIs.

2. The computer-readable storage medium of claim 1 , the operations further including comparing the workload determinism KPIs with respective overall edge node determinism metrics of the respective ones of the candidate target edge nodes to determine whether the respective ones of the candidate target edge nodes support the workload determinism KPIs.

3. The computer-readable storage medium of claim 1 , the operations further comprising calculating overall edge node determinism metrics for each of the respective ones of the candidate target edge nodes of the set of candidate target edge nodes, calculating the overall edge node determinism metrics being based on one or more edge node subsystem determinism metrics for said each of the respective ones of the candidate target edge nodes.

4. The computer-readable storage medium of claim 3 , wherein:

the one or more edge node subsystem determinism metrics for said each of the respective ones of the candidate target edge nodes include edge node subsystem determinism metrics d ij corresponding to:

a subsystem number i designating a subsystem type of said each of the respective ones of the candidate target edge nodes, where i is an integer equal to or greater than 1; and

a data point instance j of the one or more edge node subsystem determinism metrics for each subsystem number i, where j is an integer equal to or greater than 1 and;

the operations further include:

assigning a weight vector w i to subsystem determinism metrics of each subsystem number i;

for said each of the respective ones of the candidate target edge nodes, calculating a sum n j of the one or more edge node subsystem determinism metrics for each data point instance j and through all edge node subsystem determinism metrics i, wherein n j is given by:

n

j

=

Σ

i

=

1

⁢

w

i

⁢

d

ij

;

 and

performing regression on data points n j through data point instances j to determine the overall edge node determinism metrics as a function of time, d(t), for said each of the respective ones of the candidate target edge nodes.

5. The computer-readable storage medium of claim 3 , wherein the one or more edge node subsystem determinism metrics include at least one of: edge node compute subsystem determinism metrics, edge node network subsystem determinism metrics, edge node memory subsystem determinism metrics, or edge node network subsystem determinism metrics.

6. The computer-readable storage medium of claim 1 , the operations further comprising:

determining whether respective edge nodes of the edge computing system support workload computer resource requirements of the workload; and

in response to a determination that one or more of the respective edge nodes of the edge computing system support the workload compute resource requirements:

generating a node identification list for the one or more of the respective edge nodes of the edge computing system, the node identification list to identify the set of candidate target edge nodes; and

using the node identification list for the one or more of the respective edge nodes to determine whether respective ones of the candidate target edge nodes of the set of candidate target edge nodes support the workload determinism key performance indicators (KPIs).

7. The computer-readable storage medium of claim 6 , the operations further comprising:

determining compute resource requirements of the respective edge nodes from resource landscape data for the edge computing system; and

comparing the workload computer resource requirements with the compute resource requirements of the respective edge nodes to determine whether the respective edge nodes support the workload computer resource requirements.

8. The computer-readable storage medium of claim 4 , the operations further including:

in response to a determination that the one or more candidate target edge nodes include a single candidate target edge node, selecting the single candidate target edge node as the target edge node;

in response to a determination that the one or more candidate target edge nodes include a plurality of candidate target edge nodes:

applying a cost function to the plurality of candidate target edge nodes by determining a cost of deployment of the workload on each candidate target edge node of the plurality of candidate target edge nodes; and

selecting the target edge node based on determining the cost of deployment.

9. The computer-readable storage medium of claim 8 , the operations further including, in response to the determination that the one or more candidate target edge nodes include a plurality of candidate target edge nodes, selecting the target edge node as an edge node with a lowest cost of deployment.

10. The computer-readable storage medium of claim 8 , wherein selecting the target edge node includes:

determining that the workload is to be deployed in a redundant and synchronized configuration as a primary workload on a primary edge node and a secondary workload on a secondary edge node; and

based on a determination that the workload is to be deployed in a redundant and synchronized configuration, selecting, as the target edge node, the primary edge node and the secondary edge node from the plurality of candidate target edge nodes based on a rate of change of a last calculated datapoint instance of the overall edge node determinism metrics of each of the primary edge node and the secondary edge node.

11. The computer-readable storage medium of claim 10 , wherein the last calculated datapoint instance of the overall edge node determinism metrics of each of the primary edge node and the secondary edge node corresponds to a decision metric given by:

Decision Metric= d ′( t f )* n f

where d′(t f ) is a derivative of d(t f ), and d(t f ) is a value if d(t) at a last time instance t f at which d(t) has been determined.

12. The computer-readable storage medium of claim 10 , the operations further including implementing edge node state synchronization between the primary edge node and the secondary edge node by causing a storing of workload interprocess communication (IPC) calls and trace for the workload at the primary edge node and the secondary edge node.

13. An apparatus of an edge computing system to identify a target edge node for deployment of a workload thereon, the apparatus comprising:

a centralized determinism evaluator (CDE) to:

parse one or more fields of a Virtualized Industrial Function Descriptor (VIFD);

determine, based on parsing the one or more fields of the VIFD, whether respective ones of candidate target edge nodes of a set of candidate target edge nodes of the edge computing system support workload determinism key performance indicators (KPIs) of the workload; and

in response to a determination that one or more candidate target edge nodes support the workload determinism KPIs, select a target edge node from the one or more candidate edge nodes; and

an orchestrator to deploy the workload at the target edge node.

14. The apparatus of claim 13 , the CDE to compare the workload determinism KPIs with respective overall edge node determinism metrics of the respective ones of the candidate target edge nodes to determine whether the respective ones of the candidate target edge nodes support the workload determinism KPIs.

15. The apparatus of claim 13 , the CDE to further calculate overall edge node determinism metrics for each of the respective ones of the candidate target edge nodes of the set of candidate target edge nodes, calculating the overall edge node determinism metrics being based on one or more edge node subsystem determinism metrics for said each of the respective ones of the candidate target edge nodes.

16. The apparatus of claim 15 , wherein:

the one or more edge node subsystem determinism metrics for said each of the respective ones of the candidate target edge nodes include edge node subsystem determinism metrics d ij corresponding to:

a subsystem number i designating a subsystem type of said each of the respective ones of the candidate target edge nodes, where i is an integer equal to or greater than 1; and

a data point instance j of the one or more edge node subsystem determinism metrics for each subsystem number i, where j is an integer equal to or greater than 1 and;

the CDE is further to:

assign a weight vector w i to subsystem determinism metrics of each subsystem number i;

for said each of the respective ones of the candidate target edge nodes, calculate a sum n j of the one or more edge node subsystem determinism metrics for each data point instance j and through all edge node subsystem determinism metrics i, wherein n j is given by:

n

j

=

Σ

i

=

1

⁢

w

i

⁢

d

ij

;

 and

perform regression on data points n j through data point instances j to determine the overall edge node determinism metrics as a function of time, d(t), for said each of the respective ones of the candidate target edge nodes.

17. A computer-implemented method to identify a target edge node of an edge computing system for deployment of a workload on the target edge node, the method comprising:

parsing one or more fields of a Virtualized Industrial Function Descriptor (VIFD);

determining, based on parsing the one or more fields of the VIFD, whether respective ones of candidate target edge nodes of a set of candidate target edge nodes of the edge computing system support workload determinism key performance indicators (KPIs) of the workload;

in response to a determination that one or more candidate target edge nodes support the workload determinism KPIs, selecting a target edge node from the one or more candidate edge nodes; and

causing the workload to be deployed at the target edge node.

18. The method of claim 17 , the method further comprising calculating overall edge node determinism metrics for each of the respective ones of the candidate target edge nodes of the set of candidate target edge nodes, calculating the overall edge node determinism metrics being based on one or more edge node subsystem determinism metrics for said each of the respective ones of the candidate target edge nodes wherein:

the one or more edge node subsystem determinism metrics for said each of the respective ones of the candidate target edge nodes include edge node subsystem determinism metrics d ij corresponding to:

a subsystem number i designating a subsystem type of said each of the respective ones of the candidate target edge nodes, where i is an integer equal to or greater than 1; and

a data point instance j of the one or more edge node subsystem determinism metrics for each subsystem number i, where j is an integer equal to or greater than 1 and;

the method further includes:

assigning a weight vector w i to subsystem determinism metrics of each subsystem number i;

for said each of the respective ones of the candidate target edge nodes, calculating a sum n j of the one or more edge node subsystem determinism metrics for each data point instance j and through all edge node subsystem determinism metrics i, wherein n j is given by:

n

j

=

Σ

i

=

1

⁢

w

i

⁢

d

ij

;

 and

performing regression on data points n j through data point instances j to determine the overall edge node determinism metrics as a function of time, d(t), for said each of the respective ones of the candidate target edge nodes.

19. The method of claim 17 , the method further including:

in response to a determination that the one or more candidate target edge nodes include a single candidate target edge node, selecting the single candidate target edge node as the target edge node;

in response to a determination that the one or more candidate target edge nodes include a plurality of candidate target edge nodes:

applying a cost function to the plurality of candidate target edge nodes by determining a cost of deployment of the workload on each candidate target edge node of the plurality of candidate target edge nodes; and

selecting the target edge node based on determining the cost of deployment, wherein selecting the target edge node includes:

determining that the workload is to be deployed in a redundant and synchronized configuration as a primary workload on a primary edge node and a secondary workload on a secondary edge node; and

based on a determination that the workload is to be deployed in a redundant and synchronized configuration, selecting, as the target edge node, the primary edge node and the secondary edge node from the plurality of candidate target edge nodes based on the cost of deployment and on a rate of change of a last calculated datapoint instance of the overall edge node determinism metrics of each of the primary edge node and the secondary edge node; and

implementing edge node state synchronization between the primary edge node and the secondary edge node by causing a storing of workload interprocess communication (IPC) calls and trace for the workload at the primary edge node and the secondary edge node.

20. The method of claim 19 , wherein the last calculated datapoint instance of the overall edge node determinism metrics of each of the primary edge node and the secondary edge node corresponds to a decision metric given by:

Decision Metric= d ′( t f )* n f

where d′(t f ) is a derivative of d(t f ), and d(t f ) is a value if d(t) at a last time instance t f at which d(t) has been determined.

21. The method of claim 19 , the method further including:

determining whether a request for migration of the workload from the primary edge node exists; and

in response to a determination that the request exists:

recreating the workload IPC calls at the secondary edge node; and

deleting the workload IPC calls and trace from the primary edge node; and

in response to a determination that the request does not exist, deleting the workload IPC calls and trace from the primary edge node and the secondary edge node.

22. An apparatus comprising:

means for parsing one or more fields of a Virtualized Industrial Function Descriptor (VIFD);

means for determining, based on parsing the one or more fields of the VIFD, whether respective ones of candidate target edge nodes of a set of candidate target edge nodes of the edge computing system support workload determinism key performance indicators (KPIs) of the workload;

means for, in response to a determination that one or more candidate target edge nodes support the workload determinism KPIs, selecting a target edge node from the one or more candidate edge nodes; and

means for causing the workload to be deployed at the target edge node.

Assignments (1)

Continuity (1)

Cited By (10)

US 12,229,414 US 12,242,347 US 12,255,964 US 12,342,426 US 12,468,296 US 12,474,959 US 12,483,468 US 12,536,485 US 12,568,041 US 12,664,031