Flow scheduling for machine learning workloads
AssigneeMicrosoft Technology Licensing, LLC
InventorsWei BAI, Ranveer CHANDRA, Philipp Andre WITTE
Life of this patent
Estimating the term…
Abstract
The present disclosure relates to systems and methods for flow scheduling of machine learning workloads. The systems and methods use prior knowledge of the traffic patterns for machine learning workloads and the network topology to schedule the flows of the machine learning workloads. A centralized controller leverages the knowledge of the traffic patterns for the machine learning workloads to guide per-flow routing decisions for the machine learning workflows to schedule the flows across all available network paths to fully utilize network bandwidth capacity.
Claims20
A method, comprising:
- receiving, from a controller, a routing schedule for a machine learning workload that maps data packets of each flow of the machine learning workload to machines in a network using an application identifier encoded into a source port field of a packet header of the data packets;
- forwarding each flow of the machine learning workload across paths in the network to machines based on the source port;
- monitoring the network for changes;
- and
- adjusting a path of a flow of the machine learning workload in response to detecting a network change.
A network device, comprising:
- a memory to store data and instructions;
- and
- a processor operable to communicate with the memory, wherein the processor is operable to:
- receive, from a controller, a routing schedule for a machine learning workload that maps data packets of each flow of the machine learning workload to machines in a network using an application identifier encoded into a source port field of a packet header of the data packets;
- forward each flow of the machine learning workload across paths in the network to machines based on the source port;
- monitor the network for changes;
- notify the controller of a network change;
- and
- receive, from the controller, an updated routing schedule based on the network change.
Dependencies are inferred from the claim text.