Network Service Path Discovery for NVMe Fabric Hosts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computing networks, especially in NVM Express over Fabrics (NVMe-oF) systems, each host must individually perform discovery processes to determine network information and alternative paths, which affects performance and reduces bandwidth, and can delay fault response times.
Innovation Solution
A network service that identifies paths and stores path information in persistent storage, providing host-specific portions to reduce processing loads and eliminate individual host discovery processes, while detecting faults more rapidly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If each host individually performs discovery processes to determine network information and alternative paths, then hosts can access data locations and maintain fault tolerance, but processing loads on hosts increase and network bandwidth is reduced
Solution Approach 1:
A network service is introduced as an intermediary between hosts and storage devices. This service performs path discovery and management functions centrally, eliminating the need for each host to individually execute discovery processes. The network service maintains path information and provides it to hosts upon request, thereby reducing host processing loads while preserving fault tolerance capabilities through centralized path management and alternative path identification.
2Reliability
If each host individually performs discovery processes to determine network information and alternative paths, then hosts can identify alternative paths for fault tolerance, but network bandwidth is reduced due to multiple discovery operations
Solution Approach 1:
Multiple individual discovery processes executed by different hosts are merged into a single centralized discovery process performed by the network service. The network service discovers and maintains path information for all hosts collectively, eliminating redundant discovery operations across the network. This consolidation reduces network bandwidth consumption while ensuring all hosts have access to alternative path information for maintaining reliability.
3Reliability
If hosts perform individual discovery processes to update network information, then hosts can respond to path faults, but fault response time is delayed due to separate discovery operations
Solution Approach 1:
The network service performs path discovery and maintains alternative path information in advance, before faults occur. Path information is pre-computed and stored ready for immediate retrieval. When a fault occurs, hosts can quickly obtain alternative path information from the network service without waiting for new discovery processes to execute, thereby significantly reducing fault response time while maintaining the capability to respond to path faults.
Data Source
AI summary
In one implementation, a non-transitory machine-readable storage medium may store instructions that upon execution cause a processor to: determine, by a network device, path information for a plurality of paths in a computing fabric, the path information identifying alternative paths to access namespaces; determine, by the network device, a first portion of the path information that is associated with a first host; and provide, by the network device the first portion of the path information to the first host.