Systems and methods for latency-aware edge computing

Patent metadata

Field	Value
Publication number	US-11902092-B2
Application number	US-202016790582-A
Country	US
Kind code	B2
Filing date	Feb 13, 2020
Priority date	Feb 15, 2019
Publication date	Feb 13, 2024
Grant date	Feb 13, 2024

Abstract

Provided are systems, methods, and apparatuses for latency-aware edge computing to optimize network traffic. A method can include: determining network parameters associated with a network architecture, the network architecture comprising a data center and an edge data center; determining, using the network parameters, a first programmatically expected latency associated with the data center and a second programmatically expected latency associated with the edge data center; and determining, based at least in part on a difference between the first programmatically expected latency or the second programmatically expected latency, a distribution of a workload to be routed between the data center and the edge data center.

First claim

What is claimed is:

1. A device for network optimization, the device comprising:
at least one memory device that stores computer-executable instructions; and
at least one processor configured to access the memory device, wherein the processor is configured to execute the computer-executable instructions to:
identify a workload including a request for service from a source;
determine network parameters associated with a network architecture, the network architecture comprising a data center and an edge data center;
determine, using the network parameters, a first programmatically expected latency associated with the data center and a second programmatically expected latency associated with the edge data center;
determine, based at least in part on a difference between the first programmatically expected latency or the second programmatically expected latency, a distribution of the workload to be routed between the data center and the edge data center, wherein, the distribution is determined based at least in part on the difference exceeding a predetermined threshold;
partition the-request into a first portion and a second portion;
add a first header to the first portion and a second header to the second portion; and
route the first portion of the request to the data center and the second portion of the request to the edge data center based on the distribution of the workload, wherein the processor is configured to route the first portion to the data center based on the first header, and route the second portion to the edge data center based on the second header, wherein at least one of the data center or the edge data center is configured to process at least one of the first portion or the second portion of the request and generate an output to provide the service requested by the source.

2. The device of claim 1 , wherein the network parameters comprise at least one of a usage percentage, a core data center selection probability, a delay sensitivity value, a data upload amount, a data download amount, a processor usage requirement, or a virtual machine (VM) utilization.

3. The device of claim 1 , wherein the computer-executable instructions to determine the distribution of the workload include computer-executable instruction for a machine learning technique, the machine learning technique comprising at least one of a supervised machine learning technique or an unsupervised machine learning technique, the machine learning technique further comprising at least one of a long short term memory (LSTM) neural network, a recurrent neural network, a time delay neural network, or a feed forward neural network.

4. The device of claim 1 , wherein at least one of (1) a transmission rate associated with data traffic to and from the data center, (2) a transmission rate associated with data traffic to and from the edge data center, (3) a transmission rate associated with data traffic to and from a device associated with the edge data center, or (4) a transmission rate associated with data traffic to and from a device associated with the data center is throttled based on the difference.

5. The device of claim 1 , wherein the processor is further configured to label at least a first portion of the workload with a first tag, the first tag for indicating that the first portion was processed by the data center, and the processor is configured to label at least a second portion of the workload with a second tag, the second tag for indicating that the second portion was processed by the edge data center.

6. The device of claim 5 , wherein the processor is further configured to (1) receive a first completed workload associated with the first portion from the data center, and (2) receive a second completed workload associated with the second portion from the edge data center, and (3) classify, filter, or aggregate the first completed workload or the second completed workload using the first tag or second tag.

7. The device of claim 6 , wherein the processor is configured to cause to transmit at least the first completed workload or the second completed workload to a second device on the network architecture.

8. The device of claim 1 wherein the processor is configured to execute the computer-executable instructions to:
receive a first result from the data center based on processing the first portion of the workload;
receive a second result from the edge data center based on processing the second portion of the workload; and
aggregate the first result and the second result.

9. Method for network optimization, the method comprising:
identifying a workload including a request for service from a source;
determining network parameters associated with a network architecture, the network architecture comprising a data center and an edge data center;
determining, using the network parameters, a first programmatically expected latency associated with the data center and a second programmatically expected latency associated with the edge data center;
determining, based at least in part on a difference between the first programmatically expected latency or the second programmatically expected latency, a distribution of the workload to be routed between the data center and the edge data center, wherein, the distribution is determined based at least in part on the difference exceeding a predetermined threshold;
partitioning the request into a first portion and a second portion;
adding a first header to the first portion and a second header to the second portion; and
routing the first portion of the request to the data center and the second portion of the request to the edge data center based on the distribution of the workload, wherein the routing of the first portion to the data center is based on the first header, and the routing of the second portion to the edge data center is based on the second header, wherein at least one of the data center or the edge data center is configured to process at least one of the first portion or the second portion of the request and generate an output to provide the service requested by the source.

10. The method of claim 9 , wherein the network parameters comprise at least one of a usage percentage, a core data center selection probability, a delay sensitivity value, a data upload amount, a data download amount, a processor usage requirement, or a virtual machine (VM) utilization.

11. The method of claim 9 , wherein the determining of the workload is performed using a machine learning technique, the machine learning technique comprising at least one of a supervised machine learning technique or an unsupervised machine learning technique, the machine learning technique further comprising at least one of an LSTM neural network, a recurrent neural network, a time delay neural network, or a feed forward neural network.

12. The method of claim 9 , wherein the method further comprises throttling, based on the difference, at least one of (1) a transmission rate associated with data traffic to and from the data center, (2) a transmission rate associated with data traffic to and from the edge data center, (3) a transmission rate associated with data traffic to and from a device associated with the edge data center, or (4) a transmission rate associated with data traffic to and from a device associated with the data center.

13. The method of claim 9 , wherein the method further comprises labeling at least a first portion of the workload with a first tag, the first tag for indicating that the first portion was processed by the data center, and labeling at least a second portion of the workload with a second tag, the second tag for indicating that the second

Assignees

• Samsung Electronics Co Ltd

Inventors

• Zheng Qinling
• Najafabadi Ehsan
• Zaghloul Yasser

Classifications

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/09

  Supervised learning · CPC title

• H04L41/0823 Primary

  characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability (for optimising operational conditions of wireless networks H04W24/02) · CPC title

• G06F9/45558

  Hypervisor-specific management and integration aspects · CPC title