A resource scheduling method and apparatus, an electronic device, and a storage medium are provided, which are related to the technical field of system resource scheduling. The resource scheduling method comprises: monitoring whether a current system can bear a load of a target application which has triggered and entered a high-computational-power scenario, subjecting the system to resource scheduling if the system is monitored to be unable to bear the load of the target application, and running the target application in the high-computational-power scenario based on scheduled system resources.

Techniques for remote disaggregated infrastructure processing units (IPUs) are described. An apparatus described herein includes an interconnect controller to receive a transaction layer packet (TLP) from a host compute node; identify a sender and a destination from the TLP; and provide, to a content addressable memory (CAM), a key determined from the sender and the destination. The apparatus as described herein can further include core circuitry communicably coupled to the interconnect controller, the core circuitry to determine an output of the CAM based on the key, the output comprising a network address of an infrastructure processing unit (IPU) assigned to the host compute node, wherein the IPU is disaggregated from the host compute node over a network; and send the TLP to the IPU using a transport protocol.

Techniques are disclosed for allocating resources of a computing device. An operating system executing at the computing device may receive a request for the computing device to execute a task associated with an application installed at the computing device and determine a resource cost associated with executing the task. In various examples, the operating system further determines, based on the application, an amount of resources available for executing the task and schedules the task to be executed at the computing device. Responsive to determining that the amount of resources available to execute the task is sufficient given the resource cost associated with the task, the computing device may execute the task based on the schedule.

A dynamic allocator for providing platform resource candidates is disclosed. In an implementation, a platform resource allocator receives a request from a workload initiator such as, an application, for a platform resource recommendation. The platform resource allocator analyzes performance capabilities and utilization metrics of a plurality of platform resources for each of a plurality of resource. The plurality of platform resources includes one or more graphics processor units (GPUs) and one or more accelerated processing units (APUs). The platform resource allocator dynamically provides the platform resource recommendation to the workload initiator to select one or more of the plurality of platform resources to execute a workload based on the performance capabilities and utilization metrics.

An optimized service-based pipeline includes a resource manager that receives a request that includes a description of a workload from a workload initiator such as an application. The resource manager identifies runtime utilization metrics of a plurality of processing resources, where the plurality of processing resources includes at least a first graphics processing unit (GPU) and a second GPU. The resource manager determines, based on the utilization metrics and one or more policies, a workload allocation recommendation for the workload. Thus, the workload initiator can determine whether placing a workload on a particular processing resource is preferable based on runtime behavior of the system and policies established of the workload.

Techniques for heterogeneous compute instance auto-scaling with reinforcement learning (RL) are described. A user specifies a reward function that generates rewards for use with an application simulation for determining what different instance types should be added to or removed from the application as part of training a RL model. The RL model can be automatically deployed and used to monitor an application to automatically scale the application fleet using heterogenous compute instances.

A process includes determining a first degree of throttling to apply to a polling of hardware devices by a hardware processor based on a historical total utilization of the hardware processor; and determining a second degree of throttling to apply to the polling of hardware devices by the hardware processor based on a historical polling utilization of the hardware processor. The hardware processor includes, responsive to an upcoming hardware device polling cycle for the hardware processor and based on the first degree of throttling and the second degree of throttling, regulating whether the hardware processor bypasses the hardware device polling cycle or executes the hardware device polling cycle.

A method for managing a memory by an electronic device, and the electronic device, are provided. The method includes detecting a plurality of applications being executed and using the memory of the electronic device, determining a priority for each application of the plurality of applications, determining at least one page from a plurality of pages of at least one application from the plurality of applications to be dropped based on at least one priority associated with the at least one application, and dropping the at least one page from the plurality of pages of the at least one application.

Provided are a scheduling method and apparatus based on a deep learning node computation, and a storage medium. The scheduling method includes: a to-be-computed node of a preset neural network computation graph is acquired; a node type of the to-be-computed node is determined, where the node type includes a hardware computation node and a software computation node; in a case where the node type is the hardware computation node, the hardware computation node is scheduled to a first queue, and whether a hardware computing power module corresponding to the hardware computation node is occupied or not is determined; and in a case where the hardware computing power module is not occupied, the hardware computation node is input into the hardware computing power module for computing.

A method and system for processing media content in Moving Picture Experts Group (MPEG) Network Based Media Processing (NBMP) includes receiving, from an NBMP source, a first message including a workflow descriptor document corresponding to a workflow for processing the media content; obtaining, based on the workflow, a task having a task template; obtaining, based on the task, a function having a function template; and managing the processing of the media content by transmitting, to a media processing entity, a second message instructing the media processing entity to perform the function based on the task. The first message, the workflow descriptor document, the task template, the function template, and/or the second message may be used to signal a scheme for processing the media content.