HPC Fabric Interface Channel Failover Mechanism

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing multi-processor (MP) systems face challenges in handling communication channel failures, leading to delays, resource wastage, and increased recovery times in high-performance computing (HPC) clusters due to limitations in detecting failures and managing redundant channels.

Innovation Solution

Implementing a failover mechanism that transfers the state of a communication channel between windows in a hardware fabric interface device, updating mappings between memory and hardware resources to minimize client involvement and reduce resource duplication, thereby enabling swift recovery and reducing packet loss.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If multiple identical copies of compute jobs are dispatched across different computation nodes to address communication losses, then reliability is improved, but device complexity and resource usage increase

Engineering Contradiction:

Improvecommunication reliabilityVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The communication channel state is segmented into transferable components (window state, mapping information) that can be moved between different hardware fabric interface devices. This allows failover without duplicating entire compute jobs, reducing system complexity while maintaining reliability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Instead of copying entire compute jobs, the invention copies only the essential communication channel state information between windows in fabric interface devices. This selective copying reduces resource overhead while ensuring communication reliability during failover.

Inventive Principle:

Principle #26Copying

2Reliability

If multiple active communication channels are used in active/active round robin configuration, then communication reliability is improved, but device complexity and management overhead increase

Engineering Contradiction:

Improvecommunication reliabilityVSAvoidchannel management complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The invention merges multiple communication channel states into a single failover-capable window structure. By combining state transfer, mapping updates, and client access restoration into one unified mechanism, it reduces management overhead while maintaining reliability.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The fabric interface device acts as an intermediary that manages communication channel failover transparently. It mediates between the client and underlying communication channels, handling state transfer and mapping updates without requiring complex client-side management.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Device complexity

If software timers with long intervals are used to detect channel failures, then device complexity is reduced, but loss of time increases

Engineering Contradiction:

Improvedetection mechanism simplicityVSAvoidfailure detection time

Core Design Contradiction:

Device complexityVSLoss of time

Solution Approach 1:

The invention implements feedback mechanisms through window state monitoring and mapping validation that enable rapid failure detection. The system continuously monitors communication channel health and triggers failover when anomalies are detected, reducing detection time while maintaining manageable complexity.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The fabric interface device performs preliminary actions by pre-configuring multiple windows and maintaining ready-to-use mapping information. When a failure occurs, the system can immediately switch to a pre-prepared alternative channel state, reducing detection and recovery time without requiring complex real-time analysis.

Inventive Principle:

Principle #10Preliminary action

4Reliability

If additional channel resources are assigned per end-client for failover, then communication reliability is improved, but device complexity and resource overhead increase

Engineering Contradiction:

Improvecommunication reliabilityVSAvoidresource management complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The fabric interface device windows and mapping structures serve multiple functions: they act as communication channels, store state information, and provide failover capability. This multi-functionality eliminates the need for dedicated redundant resources per client, reducing overall device complexity while maintaining reliability.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The invention changes the parameter of resource allocation from dedicated redundant channels per client to shared failover-capable window states. By transforming how resources are allocated and managed, it reduces overhead while preserving communication reliability through efficient state transfer mechanisms.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS9037898B2Communication channel failover in a high performance computing (HPC) network

Publication Date: 2015.05.19 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A method, apparatus and program product implement a failover of a communication channel in a cluster fabric that transfers a state of the communication channel between windows resident in a hardware fabric interface device. The failover is desirably implemented by updating a plurality of mappings between memory resources in a host memory and hardware resources in the fabric interface device, and typically without modifying the memory resources such that involvement of a client that utilizes the communication channel in the failover is minimized or eliminated.