Proactive Power Management in Parallel Compute Nodes

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Parallel computers consume excessive power due to the large number of compute nodes operating simultaneously, leading to high electricity costs, especially during peak usage periods, and existing power management methods are either reactive or inefficient in optimizing power usage across the system.

Innovation Solution

A proactive power management system that includes a service node connected to compute nodes through an out-of-band service network, allowing for the setting of power consumption ratios for each compute node, including processor and memory power consumption, and the initiation of jobs with specific power management attributes to optimize energy use based on user instructions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If all compute nodes operate simultaneously to maximize processing throughput, then productivity is improved, but power consumption increases excessively

Engineering Contradiction:
Improveprocessing throughputVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The system dynamically adjusts the operational state of compute nodes based on job requirements and power management policies. Compute nodes can transition between active, idle, and powered-off states, allowing the system to optimize the balance between productivity and power consumption in real-time rather than operating in fixed states

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The service node proactively manages power consumption by anticipating future power needs and pre-configuring compute nodes accordingly. Power management attributes are set before job execution begins, and the system proactively transitions nodes to appropriate power states to prevent excessive power consumption before it occurs

Inventive Principle:
Principle #10Preliminary action

2Productivity

If compute nodes are kept in active state to reduce idle time, then productivity is improved, but power consumption increases

Engineering Contradiction:
Improveexecution efficiencyVSAvoidenergy wasted during idle cycles
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The system implements periodic power management where compute nodes alternate between active execution periods and idle or powered-off periods. This periodic cycling allows the system to maintain productivity during execution phases while reducing energy waste during idle phases, creating a rhythm of operation that balances both concerns

Inventive Principle:
Principle #19Periodic action

3Loss of energy

If reactive power management is used to reduce power consumption, then energy efficiency is improved, but response time increases due to delayed power adjustments

Engineering Contradiction:
Improveenergy efficiencyVSAvoidresponse time
Core Design Contradiction:
Loss of energyVSLoss of time

Solution Approach 1:

The system performs preliminary power management actions by setting power management attributes and configuring compute nodes before jobs are submitted or before power consumption issues arise. This proactive configuration eliminates the need for time-consuming reactive adjustments, as the system is already optimized when power consumption becomes a concern

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2179359B1Proactive power management in a parallel computer
Publication Date: 2013.02.27 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • EP2179359B1 patent drawingFigure 1
  • EP2179359B1 patent drawingFigure 2
  • EP2179359B1 patent drawingFigure 3

AI summary

Proactive power management in a parallel computer, the parallel computer including a service node and a plurality of compute nodes, the service node connected to the compute nodes through an out-of-band service network, each compute node including a computer processor and a computer memory operatively coupled to the computer processor. Embodiments include receiving, by the service node, a user instruction to initiate a job on an operational group of compute nodes in the parallel computer, the instruction including power management attributes for the compute nodes; setting, by the service node in accordance with the power management attributes for the compute nodes of the operational group, power consumption ratios for each compute node of the operational group including a computer processor power consumption ratio and a computer memory power consumption ratio; and initiating, by the service node, the job on the compute nodes of the operational group of the parallel computer.