Techniques for artificial intelligence capabilities at a network switch

Patent metadata

Field	Value
Publication number	US-2025097120-A1
Application number	US-202418966019-A
Country	US
Kind code	A1
Filing date	Dec 2, 2024
Priority date	Dec 28, 2018
Publication date	Mar 20, 2025
Grant date	—

Abstract

Examples include techniques for artificial intelligence (AI) capabilities at a network switch. These examples include receiving a request to register a neural network for loading to an inference resource located at the network switch and loading the neural network based on information included in the request to support an AI service to be provided by users requesting the AI service.

First claim

What is claimed is:

1 . At least one non-transitory machine readable medium storing instructions to be executed by at least one machine to be associated with a network switch, the network switch being configurable to comprise graphics processing unit (GPU) resource circuitry, data traffic processing circuitry, and management data processing circuitry, the network switch to be used in association with multiple tenants, the instructions, when executed by the at least one machine, resulting in the network switch being configured for performance of operations comprising:
receiving, at the network switch, management data and tenant service-related data, the tenant service-related data being configurable to correspond, at least in part, to artificial intelligence (AI) service requests of the multiple tenants, respective portions of the tenant service-related data being associated with respective of the multiple tenants; generating, by the management data processing circuitry, configuration data, the configuration data to be based upon the management data, the configuration data to configure the GPU resource circuitry of the network switch to implement multi-tenant services associated with the respective of the multiple tenants, the multi-tenant services to be accessed by the respective of the multiple tenants based upon the respective portions of the tenant service-related data; and routing, by the data traffic processing circuitry, the respective portions of the tenant service-related data to the GPU resource circuitry of the network switch so as to permit the respective of the multiple tenants to access, based upon the respective portions of the tenant service-related data, the multi-tenant services associated with the respective of the multiple tenants; wherein:
the data traffic processing circuitry is configurable to implement load balancing in association with the routing of the respective portions of the tenant service-related data to the GPU resource circuitry of the network switch;
the multiple tenants are to be associated with respective tenant identification data; and
the data traffic processing circuitry is configurable to implement the routing of the respective portions of the tenant service-related data to the GPU resource circuitry of the network switch based upon priority data.

2 . The at least one non-transitory machine readable medium of claim 1 , wherein:
the network switch is to be comprised in a cloud-based network.

3 . The at least one non-transitory machine readable medium of claim 2 , wherein:
the data traffic processing circuitry and/or the management data processing circuitry comprise central processing unit (CPU) core circuitry.

4 . The at least one non-transitory machine readable medium of claim 3 , wherein:
the GPU resource circuitry of the network switch comprises multiple GPU processing cores.

5 . The at least one non-transitory machine readable medium of claim 4 , wherein:
the multi-tenant services are to be implemented via one or more neural networks, one or more inference resources, neural processing, and/or tensor processing to be implemented using the GPU resource circuitry of the network switch.

6 . The at least one non-transitory machine readable medium of claim 5 , wherein:
the respective of the multiple tenants are associated with respective service agreements to be implemented, at least in part, using the network switch; the respective service agreements are associated with the multi-tenant services; and the respective tenant identification data is configurable to be associated with respective tenant billing data to be generated based upon providing of the multi-tenant services to the respective of the multiple tenants.

7 . A method to be implemented in association with a network switch, the network switch being configurable to comprise graphics processing unit (GPU) resource circuitry, data traffic processing circuitry, and management data processing circuitry, the network switch to be used in association with multiple tenants, the method comprising:
receiving, at the network switch, management data and tenant service-related data, the tenant service-related data being configurable to correspond, at least in part, to artificial intelligence (AI) service requests of the multiple tenants, respective portions of the tenant service-related data being associated with respective of the multiple tenants; generating, by the management data processing circuitry, configuration data, the configuration data to be based upon the management data, the configuration data to configure the GPU resource circuitry of the network switch to implement multi-tenant services associated with the respective of the multiple tenants, the multi-tenant services to be accessed by the respective of the multiple tenants based upon the respective portions of the tenant service-related data; and routing, by the data traffic processing circuitry, the respective portions of the tenant service-related data to the GPU resource circuitry of the network switch so as to permit the respective of the multiple tenants to access, based upon the respective portions of the tenant service-related data, the multi-tenant services associated with the respective of the multiple tenants; wherein:
the data traffic processing circuitry is configurable to implement load balancing in association with the routing of the respective portions of the tenant service-related data to the GPU resource circuitry of the network switch;
the multiple tenants are to be associated with respective tenant identification data; and
the data traffic processing circuitry is configurable to implement the routing of the respective portions of the tenant service-related data to the GPU resource circuitry of the network switch based upon priority data.

8 . The method of claim 7 , wherein:
the network switch is to be comprised in a cloud-based network.

9 . The method of claim 8 , wherein:
the data traffic processing circuitry and/or the management data processing circuitry comprise central processing unit (CPU) core circuitry.

10 . The method of claim 9 , wherein:
the GPU resource circuitry of the network switch comprises multiple GPU processing cores.

11 . The method of claim 10 , wherein:
the multi-tenant services are to be implemented via one or more neural networks, one or more inference resources, neural processing, and/or tensor processing to be implemented using the GPU resource circuitry of the network switch.

12 . The method of claim 11 , wherein:
the respective of the multiple tenants are associated with respective service agreements to be implemented, at least in part, using the network switch; the respective service agreements are associated with the multi-tenant services; and the respective tenant identification data is configurable to be associated with respective tenant billing data to be generated based upon providing of the multi-tenant services to the respective of the multiple tenants.

13 . Circuitry to implement a network switch, the network switch being configurable to comprise graphics processing unit (GPU) resource circuitry, the network switch to be used in association with multiple tenants, the circuitry to implement the network switch comprising:
circuitry to receive, at the network switch, management data and tenant service-related data, the tenant service-related data being configurable to correspond, at least in part, to artificial intelligence (AI) service requests of the multiple tenants, respective portions of the tenant service-related data being associated with respective of the multiple tenants; management

Assignees

• Intel Corp

Inventors

• Guim Bernat Francesc
• Prabhakaran Suraj
• Doshi Kshitij A
• Ganesh Brinda
• Verrall Timothy

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0442

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

• G06N3/0495

  Quantised networks; Sparse networks; Compressed networks · CPC title

• H04L41/0816

  the condition being an adaptation, e.g. in response to network events · CPC title

• H04L41/5051

  Service on demand, e.g. definition and deployment of services in real time · CPC title