International Business Machines Corporation

Intel Corporation

Globalfoundries Inc.

Lenovo (singapore) Pte. Ltd.

Heather Lynn Hanson

Karthick Rajamani

Malcolm Scott Ware

Freeman Leigh Rawson, Iii

Charles Robert Lefurgy

Malcolm Scott Allen-Ware

Todd Jon Rosedahl

Elmootazbellah Nabil Elnozahy

Alan J Drake

John Bruce Carter

Guillermo Jesus Silva

Timothy Glen Hallett

Jordan Ross Keuseman

Srinivasan Ramani

Venkat Rajeev Indukuru

Francis Patrick O'Connell

Ken Van Vu

James Lyle Peterson

David John Palframan

Alper Buyuktosunoglu

Pradip Bose

Michael Stephen Floyd

Wei T Huang

Hans Mikael Jacobson

Richard Loise Arndt

Juan C Rubio

Wael R El-Essawy

Thomas W Keller, Jr

Augusto Javier Vega

Bradley Gene Burgess

Arthur M Ryan

Michael L Choate

Mark D Nicol

Chandrakant Pandya

Freeman L Rawson

Todd D Erdner

Michael J Borsch

Malcom S Allen-Ware

A distributed power management system is configured determine a node power consumption of a node during a first time interval. The system can determine a node power cap. The system can determine a proportional component power budget for a component of the node based, at least in part, on the node power consumption and a component power consumption. The system can determine a power budget for the component for a second time interval based, at least in part on the proportional component power budget.

Distributed power budgeting

A processor system tracks, in at least one counter, a number of cycles in which at least one execution unit of at least one processor core is idle and at least one thread of the at least one processor core is waiting on at least one off-core memory access during run-time of the at least one processor core during an interval comprising multiple cycles. The processor system evaluates an expected performance impact of a frequency change within the at least one processor core based on the current run-time conditions for executing at least one operation tracked in the at least one counter during the interval.

Tracking pipelined activity during off-core memory accesses to evaluate the impact of processor core frequency changes

A mechanism is provided for optimizing energy efficiency in a set of processor cores while maintaining application performance for a set of applications. A quality of service (QoS) level is received for one or more active applications in the set of applications and state information associated with each processor core in the set of processor cores is identified. Responsive to the QoS level and the state information indicating an action to be implemented, a change is implemented to reduce power utilization by one or more processor cores in the set of processor cores in the data processing system, where the change is via at least one of dynamic frequency scaling, dynamic voltage scaling, or core folding.

Energy efficient optimization in multicore processors under quality of service (QoS)/performance constraints

According to an aspect, power management of a multi-core processing system includes determining workload characteristics in the multi-core processing system. A power adjustment scenario is identified based on the workload characteristics. A predetermined actuation order for at least two power adjustment actuators is identified based on the power adjustment scenario. Based on the predetermined actuation order, it is determined whether there is an adequate adjustment capacity for a power adjustment action associated with one of the at least two power adjustment actuators. The power adjustment action is initiated based on the predetermined actuation order and determining that the adequate adjustment capacity is available.

Power management for multi-core processing systems

A distributed power management computer program product is configured to collect power consumption data that indicates power consumption by at least a plurality of the components of a node. The program code can be configured to provide, to each of a plurality of controllers associated with a respective one of the plurality of components, the power consumption data. The program code can be configured to determine a node power consumption. The program code can be configured to determine a power differential as a difference between the node power consumption and an upper power consumption threshold of the node. The program code can be configured to determine a proportion of the node power consumption consumed by a first component. The program code can be configured to compute a local power budget for the first component.

Embodiments include collecting, from each of a plurality of controllers of a node having a plurality of components, component power consumption. Each of the plurality of controllers is associated with one or more of the components. The component power consumptions are provided to the controllers. A node power consumption for the node is determined based, at least in part, on the component power consumption. The power cap is determined for the plurality of components. A power differential power is determined as a difference between the node power consumption and the power cap for the plurality of components. A proportion of the node power consumption consumed by the component is determined based on the component power consumption of the component. A local power budget is computed for the component based, at least in part, on the power differential and the proportion of the node power consumption consumed by the component.

A method, system, and computer program product for reducing power and energy consumption in a server system with multiple processor cores is disclosed. The system may include an operating system for scheduling user workloads among a processor pool. The processor pool may include active licensed processor cores and inactive unlicensed processor cores. The method and computer program product may reduce power and energy consumption by including steps and sets of instructions activating spare cores and adjusting the operating frequency of processor cores, including the newly activated spare cores to provide equivalent computing resources as the original licensed cores operating at a specified clock frequency.

Method of exploiting spare processors to reduce energy consumption

A mechanism is provided for priority-based power capping. A power management controller identifies a set of priorities for a set of partitions of the data processing system. The power management controller determines whether a measured power of the data processing system exceeds a power cap for the data processing system. Responsive to the measured power exceeding the power cap, the power management controller sends a set of commands to a set of component actuators to adjust one or more of a set of operation parameters for a set of components associated with the set of partitions using the set of priorities. The set of component actuators adjust the one or more of the set of operational parameters associated with the set of component in order to reduce a power consumption of the data processing system.

Priority-based power capping in data processing systems

A method for a framework for scheduling tasks in a multi-core processor or multiprocessor system is provided in the illustrative embodiments. A thread is selected according to an order in a scheduling discipline, the thread being a thread of an application executing in the data processing system, the thread forming the leader thread in a bundle of threads. A value of a core attribute in a set of core attributes is determined according to a corresponding thread attribute in a set of thread attributes associated with the leader thread. A determination is made whether a second thread can be added to the bundle such that the bundle including the second thread will satisfy a policy. If the determining is affirmative, the second thread is added to the bundle. The bundle is scheduled for execution using a core of the multi-core processor.

Framework for scheduling multicore processors

Data center environmental sensing is provided by a measurement system that detects environmental events from inputs received from a plurality of movable sensors. The sensors are moved in response to detection of an event to a region of the data center associated with the event, providing increased spatial resolution of the measurement in the region of the event. Events such as leakage between hot and cold aisles of a data center can be detected by the system, which may use one or more movable devices that can be moved around, between and over equipment, to carry multiple sensors toward the source of the event, providing both diagnostic and detailed environmental information.

Sensor-swarm environmental event detection

A mechanism is provided for scheduling application tasks. A scheduler receives a task that identifies a desired frequency and a desired maximum number of competing hardware threads. The scheduler determines whether a user preference designates either maximization of performance or minimization of energy consumption. Responsive to the user preference designating the performance, the scheduler determines whether there is an idle processor core in a plurality of processor cores available. Responsive to no idle processor being available, the scheduler identifies a subset of processor cores having a smallest load coefficient. From the subset of processor cores, the scheduler determines whether there is at least one processor core that matches desired parameters of the task. Responsive to at least one processor core matching the desired parameters of the task, the scheduler assigns the task to one of the at least one processor core that matches the desired parameters.

Optimizing energy consumption and application performance in a multi-core multi-threaded processor system

Growing community of inventors

Austin, TX, United States of America

Heather Lynn Hanson