Leaf-Switch CDC Placement for Low-Latency NVMe Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In network fabrics with a CLOS topology, NVMe endpoints experience excessive latency, delay, and jitter due to centralized placement of CDC services, impacting overall network performance.
Innovation Solution
Implement push and pull registration mechanisms to dynamically place CDC services closer to NVMe endpoints, utilizing leaf switches with distributed services and direct discovery controllers, ensuring predictable behavior in connection scale, delay, and jitter.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If CDC services are placed centrally on leaf switches, then device complexity is reduced and ease of operation is improved, but latency and jitter increase due to distance from NVMe endpoints
Solution Approach 1:
The patent segments the centralized CDC service into distributed service instances deployed across multiple leaf switches. Each leaf switch runs local service instances that handle discovery requests from NVMe endpoints in its vicinity, eliminating the need for all requests to traverse to a central service location. This segmentation reduces latency and jitter while maintaining operational simplicity through automated service placement.
Solution Approach 2:
The patent introduces a new dimension of service placement by deploying CDC services at the edge of the network fabric on leaf switches rather than concentrating them at the center. This dimensional shift from centralization to distribution across the network topology enables local service delivery, reducing transmission distance and improving response time for NVMe endpoint discoveries.
2Loss of time
If CDC services are placed closer to NVMe endpoints, then latency and jitter are reduced, but device complexity increases due to distributed service placement
Solution Approach 1:
The patent implements self-service through automated service placement mechanisms where the system autonomously determines optimal locations for CDC service instances on leaf switches based on NVMe endpoint distributions. The service placement and management is handled automatically without requiring manual configuration or complex administrative overhead, thus reducing perceived device complexity while achieving low-latency service delivery.
Solution Approach 2:
The patent changes the placement parameter of CDC services from centralized to distributed across multiple leaf switches. By adjusting this spatial parameter and deploying service instances at different locations in the network fabric, the system optimizes for reduced latency and jitter while the automation framework manages the complexity of distributed deployment.
3Ease of operation
If centralized CDC service is used, then service management is simplified, but connection scale handling becomes problematic due to resource constraints on single leaf switch
Solution Approach 1:
The patent segments the CDC service functionality across multiple leaf switches, with each switch hosting service instances that handle a portion of the total connection load. This segmentation distributes the processing burden, enabling the system to scale connections across multiple service instances rather than being limited by the resource capacity of a single leaf switch, while maintaining simplified management through automated service orchestration.
Solution Approach 2:
The patent creates universal service instances that can handle discovery requests from multiple NVMe endpoints across the network fabric. Each leaf switch with deployed service instances can serve multiple endpoints locally, providing multi-functional capability that scales with connection demands while maintaining consistent service management practices.
Data Source
AI summary
In a network fabric, such as a network having CLOS topology, non-volatile memory express (NVMe®) endpoints may be connected to one of centralized discovery controller (CDC) distributed services placed among multiple leaf switches, which may have resource constraints. Connection scale, delay, and jitter may occur if the CDC distributed services are not placed on leaf switches close to NVMe endpoints served by the CDC distributed services. System and method embodiments are disclosed for placement of CDC services on a switching network fabric close to endpoints that are served by the CDC services. The CDC services may be placed on desired leaf switches close to endpoints.