Logical Processor Distribution Across Physical Chips

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In computer systems with over-configured logical partitions, there is a performance concern regarding the balanced distribution of shared logical processors across physical processor chips, as surplus processors must be spread to adhere to the architectural rule that each logical processor requires a backing physical processor, while minimizing impact on other running partitions.

Innovation Solution

A method where the computer assigns logical processors to the physical processor chip with the greatest logical processor entitlement for a particular partition first, and then distributes remaining processors sequentially to other chips, ensuring balanced distribution and reducing contention for additional resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If logical processors are assigned to physical processor chips in sequential order, then the distribution process is simple and fast, but the load balance across chips deteriorates

Engineering Contradiction:
Improveprocessor distribution speedVSAvoidload balance across chips
Core Design Contradiction:
ProductivityVSStability of the object's composition

Solution Approach 1:

The system performs preliminary classification of logical processors into groups based on their affinity to physical processor chips before assignment. This pre-grouping ensures that processors with similar characteristics are assigned to the same chip, achieving load balance before the actual distribution occurs, thus resolving the contradiction between distribution speed and load balance.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system applies different assignment strategies to different physical processor chips based on their specific characteristics and current load states. Each chip receives logical processors tailored to its capabilities, rather than a uniform sequential assignment, thereby optimizing both distribution efficiency and load balance across heterogeneous chips.

Inventive Principle:
Principle #3Local quality

2Productivity

If logical processors are concentrated on chips with highest entitlement first, then resource utilization efficiency improves, but contention for additional resources increases

Engineering Contradiction:
Improveresource utilization efficiencyVSAvoidresource contention
Core Design Contradiction:
ProductivityVSObject-generated harmful factors

Solution Approach 1:

The system dynamically adjusts the assignment of logical processors to physical chips based on real-time entitlement states and load conditions. As chips reach their entitlement limits or become saturated, the system automatically redirects subsequent processors to alternative chips, maintaining high utilization while preventing resource contention through adaptive redistribution.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system initially assigns processors aggressively to high-entitlement chips to maximize resource utilization, then applies corrective redistribution when entitlement limits are approached. This partial excessive action followed by correction achieves high overall utilization while managing contention through staged assignment and rebalancing.

Inventive Principle:
Principle #16Partial or excessive action

3Stability of the object's composition

If logical processors are evenly distributed across all chips, then load balance improves, but the complexity of the distribution algorithm increases

Engineering Contradiction:
Improveload balanceVSAvoiddistribution algorithm complexity
Core Design Contradiction:
Stability of the object's compositionVSDevice complexity

Solution Approach 1:

The system segments the distribution process into distinct phases: classification of processors into affinity groups, initial assignment to preferred chips, and subsequent rebalancing only where needed. This segmentation simplifies the overall algorithm by breaking it into manageable steps, achieving load balance without requiring complex global optimization across all chips simultaneously.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS11782872B2Distribution of over-configured logical processors
Publication Date: 2023.10.10 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US11782872B2 patent drawing
  • US11782872B2 patent drawing
  • US11782872B2 patent drawing

AI summary

Logical processor distribution across physical processors is provided. A set of logical processors of a number of logical processors defined for a particular logical partition of a plurality of active logical partitions is assigned to a physical processor chip having a greatest logical processor entitlement for the particular logical partition until no more logical processors can be assigned to that physical processor chip based on a logical processor entitlement of that physical processor chip being exhausted. Remaining logical processors of the number of logical processors defined for the particular logical partition are assigned to other physical processor chips of a plurality of physical processor chips assigned to the particular logical partition until all of the remaining logical processors have been assigned to a physical processor chip.