Hewlett Packard Enterprise Development LP

Diman Zad Tootaghaj

Puneet Sharma

Faraz Ahmed

Junguk Cho

Anu Mercian

Michael Zayats

Lianjie Cao

Vivek Adarsh

Aditya Dhakal

Mehrnaz Sharifian

Yunming Xiao

Systems and methods are provided for improving autotuning procedures using stateless processing with a remote key-value store. For example, the system can implement a task launcher, a scheduler, and an agent to launch, schedule, and execute decomposed autotuning stages, respectively. The scheduling policy implemented by the scheduler may perform operations beyond a simple scheduling policy (e.g., a FIFO-based scheduling policy), which produces a high queuing delay. Compared to the traditional systems, by leveraging autotuning specific domain knowledge, queueing delay is reduced and resource utilization is improved.

Context-aware and stateless deep learning autotuning framework

Example implementations relate to scheduling of jobs for a plurality of graphics processing units (GPUs) providing concurrent processing by a plurality of virtual GPUs. According to an example, a computing system including one or more GPUs receives a request to schedule a new job to be executed by the computing system. The new job is allocated to one or more vGPUs. Allocations of existing jobs are updated to one or more vGPUs. Operational cost of operating the one or more GPUs and migration cost of allocating the new job are minimized and allocations of the existing jobs on the one or more vGPUs is updated. The new job and the existing jobs are processed by the one or more GPUs in the computing system.

Scheduling jobs on graphical processing units

A device and corresponding method are provided determining a consumed computing capacity of a first networking device exceeds the threshold for total capacity for processing monitoring data for a monitoring metric. An optimization engine determines a second networking device with unused computing capacity sufficient for processing the monitoring data generated by the first networking device. The optimization engine automatically moves the monitoring data for the monitoring metric generated by the first networking device to the second networking device and causes the second networking device to process the monitoring data.

Distributed network monitoring

In certain embodiments, a method includes receiving, at an interface of a Smart network interface card (SmartNIC) of a computing device, via a network, a network data unit; processing, by a data allocator of a SmartNIC subsystem of the SmartNIC, the network data unit to make a determination that data included in the network data unit is intended for processing by an accelerator of the computing device, wherein the accelerator is configured to execute a machine learning algorithm; storing, by the data allocator and based on the determination, the data in a local buffer of the SmartNIC subsystem; identifying, by the data allocator, a memory resource associated with the accelerator; and transferring the data from the local buffer to the memory resource.

Smart network interface card control plane for distributed machine learning workloads

Example implementations relate to consensus protocols in a stretched network. According to an example, a distributed system includes continuously monitoring network performance and/or network latency among a cluster of a plurality of nodes in a distributed computer system. Leadership priority for each node is set based at least in part on the monitored network performance or network latency. Each node has a vote weight based at least in part on the leadership priority of the node. Each node's vote is biased by the node's vote weight. The node having a number of biased votes higher than a maximum possible number of votes biased by respective vote weights received by any other node in the cluster is selected as a leader node.

Leader election in a distributed system based on node weight and leadership priority based on network performance

Embodiments described herein are generally directed to an edge-CaaS (eCaaS) framework for providing life-cycle management of containerized applications on the edge. According to an example, declarative intents are received indicative of a use case for which a cluster of a container orchestration platform is to be deployed within an edge site that is to be created based on infrastructure associated with a private network. A deployment template is created by performing intent translation on the declarative intents and based on a set of constraints. The deployment template identifies the container orchestration platform selected by the intent translation. The deployment template is then executed to deploy and configure the edge site, including provisioning and configuring the infrastructure, installing the container orchestration platform on the infrastructure, configuring the cluster within the container orchestration platform, and deploying a containerized application or portion thereof on the cluster.

Deployment and configuration of an edge site based on declarative intents indicative of a use case

Example implementations relate to edge acceleration by offloading network dependent applications to a hardware accelerator. According to one embodiment, queries are received at a cluster of a container orchestration platform. The cluster includes a host system and a hardware accelerator, each serving as individual worker machines of the cluster. The cluster further includes multiple worker nodes and a master node executing on the host system or the hardware accelerator. A first worker node executes on the hardware accelerator and runs a first instance of an application. A distribution of the queries is determined among the worker machines based on a queuing model that takes into consideration the respective compute capacities of the worker machines. Responsive to receipt of the queries by the host system or the hardware accelerator, the queries are directed to the master node or one of the worker nodes in accordance with the distribution.

Directing queries to nodes of a cluster of a container orchestration platform distributed across a host system and a hardware accelerator of the host system

An example network orchestrator includes processing circuitry and a memory. The memory includes instructions that cause the network orchestrator to receive network probe information including delay times of network probes associated with a set of flows between devices. The instructions further cause the network orchestrator to generate a correlation matrix including correlations representing shared congested links between pairs of flows. The instructions further cause the network orchestrator to for each flow of the set of flows, determine a routing solution optimized for the each flow and select a total minimum cost solution from the determined routing solutions.

Heuristic-based SD-WAN route reconfiguration

An example client device includes processing circuitry and a memory including instructions that, when executed by the processing circuitry, cause the client device to undertake certain actions. Certain instructions cause the device to periodically measure active network performance data for a network, calculate expected rewards for the plurality of entry points, select an expected best entry point based on the expected rewards, route data to the selected entry point, measure passive network performance data for the selected entry point, and update a reinforcement learning algorithm, based in part on the measured passive network performance data.

Dynamic network service data routing based on measured network performance data

Example implementations relate to a proactive auto-scaling approach. According to an example, a machine-learning prediction model is trained to forecast future serverless workloads during a window of time for an application running in a public cloud based on past serverless workload information associated with the application by performing a training process. During the window of time, serverless workload information associated with the application is monitored. A future serverless workload is predicted for the application at a future time within the window, based on the machine learning prediction model. Prior to the future time, containers within the public cloud executing the application are pre-warmed to accommodate the predicted future serverless workload by issuing fake requests to the application to trigger auto-scaling functionality implemented by the public cloud.

Proactively accomodating predicted future serverless workloads using a machine learning prediction model

Example implementations relate to a proactive auto-scaling approach. According to an example, a target performance metric for an application running in a serverless framework of a private cloud is received. A machine-learning prediction model is trained to forecast future serverless workloads during a window of time for the application based on historical serverless workload information. The serverless framework is monitored to obtain serverless workload observations for the application. A future serverless workload for the application at a future time is predicted by the trained machine learning prediction model based on workload observations. A feedback control system is then used to output a new number of replicas based on a current value of the performance metric, the target performance metric and the predicted future serverless workload. Finally, the serverless framework is caused to scale and pre-warm a number of replicas supporting the application to the new number.

Growing community of inventors

Palo Alto, CA, United States of America

Diman Zad Tootaghaj