Kernel Thread I/O Re-prioritization Offload

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional VTIOL processing is hindered by the offloading of numerous I/O operations, leading to slowdowns and delays in processing, which negatively impacts system performance, especially when handling high-priority tasks.

Innovation Solution

A virtual memory manager (VMM) with VTIOLR capability, where pending I/O operations are offloaded to kernel threads and re-prioritized by a VTIOLR thread, using specific schemes and tunable conditions to optimize processing and reduce blocking cycles.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If I/O operations are offloaded to kernel threads, then processing throughput is improved, but processing delays and blocking cycles increase

Engineering Contradiction:
ImproveI/O processing throughputVSAvoidProcessing delay
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent segments the I/O processing workload by introducing separate kernel threads (first kernel threads for offloaded operations, second kernel thread for re-prioritized operations) that handle different portions of I/O operations independently. This segmentation allows parallel processing of I/O operations, improving throughput while managing delays through dedicated processing channels.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic re-prioritization where the re-prioritization controller dynamically transfers I/O operations between kernel threads based on priority levels and processing conditions. This dynamic adjustment allows the system to adapt to varying workloads, ensuring high-priority operations receive timely attention while maintaining overall throughput.

Inventive Principle:
Principle #15Dynamics

2Ease of operation

If numerous I/O operations are offloaded, then system responsiveness is improved, but processing slowdowns occur

Engineering Contradiction:
ImproveSystem responsivenessVSAvoidProcessing speed
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The patent applies partial action by having the VTIOLR thread handle only a portion of re-prioritization tasks rather than all operations. The system selectively re-prioritizes based on defined conditions (e.g., priority levels, operation types), which maintains responsiveness for critical operations while avoiding the overhead of re-prioritizing every single I/O operation, thus preventing processing slowdowns.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The re-prioritization controller acts as an intermediary between the VTIOL threads and the I/O operations. It mediates the transfer of operations between kernel threads, managing the balance between responsiveness and processing speed by selectively transferring only those operations that meet specific criteria, thereby avoiding unnecessary re-prioritization overhead.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Device complexity

If I/O operations are handled at interrupt level, then processing simplicity is maintained, but blocking cycles and performance degradation occur

Engineering Contradiction:
ImproveProcessing structure simplicityVSAvoidSystem performance
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent segments the interrupt-level processing by introducing separate kernel threads that handle different categories of I/O operations. The first kernel threads handle offloaded operations while the second kernel thread handles re-prioritized operations, dividing the processing workload to maintain simplicity in each thread while improving overall system performance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The re-prioritization controller serves as an intermediary layer between the simple interrupt-level handling and the more complex multi-threaded processing. It translates and manages I/O operations, transferring them between appropriate kernel threads based on priority and processing requirements, thus maintaining the simplicity of interrupt-level processing while enabling enhanced performance through structured multi-threaded execution.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS10235202B2Thread interrupt offload re-prioritization
Publication Date: 2019.03.19 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US10235202B2 patent drawing
  • US10235202B2 patent drawing
  • US10235202B2 patent drawing

AI summary

A computing system is provided and includes first and second computing resources defined, during system initialization, as first kernel threads and a second kernel thread with which the first kernel threads are operably associated, a memory manager and a re-prioritization controller. The memory manager is configured to handle a portion of pending input/output (I/O) operations at an interrupt level and to offload a remainder of the pending I/O operations to the first kernel threads according to an offload condition whereby the offloaded I/O operations are queued according to a first scheme. The re-prioritization controller is configured to transfer a portion of the offloaded I/O operations from the first kernel threads to the second kernel thread according to a transfer condition whereby the transferred I/O operations are re-prioritized according to a second scheme.