公开(公告)号：US20250126552A1

公开(公告)日：2025-04-17

申请号：US18664101

申请日：2024-05-14

Applicant: Cisco Technology, Inc.

Inventor： Marisol Palmero Amador ,  Snezana Mitrovic ,  Derek W. Engi ,  Gonzalo A. Salgueiro ,  Joel Richard Goergen

IPC: H04W52/02 ,  H04W24/04

Abstract: Devices, systems, methods, and processes for feature level power calibration are described herein. Network devices include sensors that generate sensor readings indicative of various device parameters. A calibration logic utilizes the sensor readings and feature permutations associated with the sensor readings to predict a feature level power consumption for all features of the network device. The calibration logic then applies a calibration factor to the predicted feature level power consumption and obtains an actual feature level power consumption. Using the actual feature level power consumption, the calibration logic determines an actual power consumption for feature licenses of the network device. The feature and feature license level power consumption is utilized for determining which features or feature licenses can be deactivated when the device power consumption is outside a threshold limit. Such dynamic deactivation ensures that the network device accurately meets the sustainability goals.

公开(公告)号：US20240155500A1

公开(公告)日：2024-05-09

申请号：US18344107

申请日：2023-06-29

Applicant: Cisco Technology, Inc.

Inventor： Derek W. Engi ,  Gonzalo A. Salgueiro ,  Marisol Palmero Amador ,  Md Atiqur Rahman ,  Pascal Thubert

IPC: H04W52/02

CPC classification number: H04W52/0264 ,  H04W52/0251

Abstract: Described herein are devices, systems, methods, and processes for managing power consumption in network nodes by dynamically disabling and reenabling features based on real-time power consumption monitoring and historical data analysis. A machine learning model is trained using historical sensor data and historical feature data to predict power consumption and derive feature-to-power association data. The power budget is determined based on sustainability goals. Real-time power consumption is monitored, and features are disabled or reenabled based on their priorities and power consumption levels to maintain the power budget. The machine learning model is validated and updated using new historical data to improve its prediction accuracy and adaptability. The feature-to-power association data is distributed to network nodes and management systems for power management purposes. These devices, systems, methods, and processes enable efficient power management in network nodes while maintaining optimal network performance and contributing to sustainability goals.

公开(公告)号：US20240154826A1

公开(公告)日：2024-05-09

申请号：US18344561

申请日：2023-06-29

Applicant: Cisco Technology, Inc.

Inventor： Pascal Thubert ,  Gonzalo A. Salgueiro ,  Derek W. Engi ,  Marisol Palmero Amador

IPC: H04L12/12 ,  H04L12/46 ,  H04L45/745

CPC classification number: H04L12/12 ,  H04L12/4633 ,  H04L45/74591

Abstract: Described herein are devices, systems, methods, and processes for intelligently managing power consumption in a network by allocating a power budget for packet processing. The power budget can be allocated based on criticality and/or the trust level of the flow. A network device may determine which subsets of features can be executed within the power budget for specific flows. Network devices can signal their capability to run features based on power consumption and adherence to the power budget, allowing for cooperative end-to-end power-based decision-making and policy enforcement. Network devices unable to run all features can select a subset of the features within their power budget and a viable path where other network devices can execute the missing features. Source route information can be added to indicate the path and missing features to be executed by network devices down the segment routing path.