Physical Layer Device for Server Network Redundancy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face downtime and service disruptions due to single points of failure in server connectivity, which can lead to lost revenue and customer dissatisfaction, especially since existing solutions like NIC teaming and multi-chassis link aggregation are difficult to implement and may result in traffic forwarding issues.
Innovation Solution
Implementing a method that connects each server to two different Tier-0 network elements through diverse routes using physical layer devices with cross-point switch capabilities and active DAC Y-cables, allowing for quick diversion of traffic to an alternate path in case of failures, and using out-of-band control plane signals for coordinated switchover.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If NIC teaming and multi-chassis link aggregation are implemented, then network resiliency is improved, but device complexity and ease of operation deteriorate due to implementation difficulties and traffic forwarding issues
Solution Approach 1:
The patent introduces a physical layer device (PLD) as an intermediary component that sits between the server NIC and the network switches. This PLD with cross-point switch capability handles the complexity of path selection and failover, while the server itself only needs a single NIC configuration. The PLD mediates the connection by monitoring link status and automatically switching traffic paths without requiring complex software configuration on the server or network devices.
Solution Approach 2:
The patent replaces the software-based NIC teaming mechanism with a hardware-based physical layer solution. Instead of relying on operating system-level NIC teaming software that requires complex configuration and management, the solution uses a physical layer device with hardware-based link monitoring and automatic path switching. This mechanical/hardware approach simplifies the system by moving the intelligence from software to hardware.
2Reliability
If a second NIC is provided to each server for redundancy, then network resiliency is improved, but cost increases due to custom NICs with complex acceleration logic
Solution Approach 1:
The patent merges the functions of multiple NICs into a single physical interface by placing a physical layer device with cross-point switch capability at the network side. Instead of requiring two separate NICs in the server, the solution combines the redundant path functionality into a single PLD that handles multiple logical connections through one physical interface, reducing hardware requirements while maintaining redundancy.
Solution Approach 2:
The physical layer device acts as an intermediary that provides the redundancy functionality without requiring the server to have multiple NICs. The PLD sits between the single server NIC and the network switches, creating virtual multiple paths through hardware switching capability, thereby eliminating the need for expensive custom dual NIC configurations in each server.
3Reliability
If diverse routing paths are implemented, then tolerance to failures is improved, but device complexity increases due to physical layer devices and active DAC Y-cables
Solution Approach 1:
The patent implements preliminary action by pre-configuring the physical layer device with multiple established paths before any failure occurs. The PLD is pre-programmed with the cross-point switching matrix and has backup paths ready and tested in advance. When a failure is detected, the device can immediately switch to a pre-established alternate path without needing to dynamically calculate or configure new routes, thus handling complexity in advance rather than during failure events.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In a group of servers, a link fault to one of the servers is detected. In response to detecting the link fault, network traffic that uses the faulty link is distinguished and the distinguished network traffic is forwarded. Applicable backup links are activated, and the distinguished traffic is redirected using an activated backup link for the faulty link.