Systems and methods for managing artificial intelligence (ai) accelerators in ai appliances
AssigneeDell Products L.P.
InventorsRobert C. Hernandez, Todd Erick Swierk
Life of this patent
Estimating the term…
Abstract
Systems and methods for managing Artificial Intelligence (AI) accelerators in AI appliances are described. In an illustrative, non-limiting embodiment, an Information Handling System (IHS) may include: a processor; and a memory coupled to the processor, the memory having program instructions stored thereon that, upon execution, cause the IHS to: receive, from a plurality of AI appliances over a network, data related to a plurality of AI appliances, where the data indicates whether a client IHS is directly coupled to an AI appliance; and select one of a plurality of AI appliances for execution of a workload on behalf of another client IHS based, at least in part, upon the data.
Claims20
An Information Handling System (IHS), comprising:
- a processor;
- and
- a memory coupled to the processor, the memory having program instructions stored thereon that, upon execution, cause the IHS to:
- receive, from a plurality of Artificial Intelligence (AI) appliances over a network, data related to a plurality of AI appliances, wherein the data indicates whether a client IHS is directly coupled to an AI appliance;
- and
- select one of a plurality of AI appliances for execution of a workload on behalf of another client IHS based, at least in part, upon the data.
A docking station, comprising:
- a Systems-on-Chip (SoC);
- an Artificial Intelligence (AI) accelerator coupled to the SoC;
- and
- a memory coupled to, or integrated into the SoC, the memory having program instructions stored thereon that, upon execution by the SoC, cause the docking station to:
- in response to a coupling to a client Information Handling System (IHS) via a local port, enable enumeration of the AI accelerator by the client IHS via a crossbar;
- and
- in response to a request by the client IHS to release the AI accelerator, disable enumeration of the AI accelerator via the crossbar.
A method, comprising:
- receiving, at an orchestrator from a plurality of Artificial Intelligence (AI) appliances over a network, telemetry data related to the plurality of AI appliances, wherein each of the plurality of AI appliances comprises a plurality of discrete AI accelerators, and wherein the telemetry data indicates whether a client IHS is directly coupled to an AI appliance;
- and
- balancing workloads across the plurality of discrete AI accelerators based, at least in part, upon whether client IHSs are directly coupled to the AI appliances.
Dependencies are inferred from the claim text.