Ring-Configured Server Networking Without Top-of-Rack Switches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High availability networks in data centers require significant space, power, and cooling resources due to the use of top-of-rack switches, which also incur additional maintenance and equipment costs.
Innovation Solution
Implementing network hardware components (NHCs) with integrated routing, switching, and encryption capabilities within servers, configured in a ring topology, eliminating the need for top-of-rack switches and offloading routing and encryption functions from server processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If top-of-rack switches are used to provide high availability networks, then network reliability is improved, but space requirements, power consumption, and cooling resources increase
Solution Approach 1:
The patent combines routing and switching functionality directly into the server's network interface card, eliminating the need for separate top-of-rack switches. This merging of functions reduces the number of discrete network hardware components required in the data center, thereby reducing space requirements while maintaining network reliability through the virtualized routing capability embedded in each server.
2Reliability
If top-of-rack switches are deployed for high availability, then network reliability is improved, but power consumption and cooling resources increase
Solution Approach 1:
By integrating routing functionality into the network interface card and eliminating separate top-of-rack switches, the patent reduces the total power consumption of the network infrastructure. The virtualized routing operations consume significantly less power compared to dedicated hardware switches, while network availability is maintained through the redundant path capabilities provided by the ring configuration.
3Reliability
If multiple top-of-rack switches are used to eliminate single points of failure, then network reliability is improved, but equipment costs and maintenance complexity increase
Solution Approach 1:
The patent extracts the routing functionality from traditional network infrastructure components and relocates it directly into the server's network interface card. This extraction eliminates the need for multiple top-of-rack switches and their associated complex interconnections, thereby reducing equipment costs and maintenance complexity while preserving fault tolerance through the distributed routing architecture.
4Adaptability or versatility
If routing functionality is implemented on servers, then network flexibility is improved, but server computing resources are consumed
Solution Approach 1:
The patent introduces a virtualized routing layer that operates as an intermediary between the server's application processing and the physical network interface. This virtual router, implemented through software-defined networking principles, provides flexible routing capabilities without significantly impacting server computing resources by operating at the network interface level rather than consuming application-level processing power.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
The technology is directed to methods and systems in which network hardware components and their associated functionality are incorporated into servers, the network hardware components functioning to couple the servers to one or more networks. The network hardware components include virtual routers and are configured to form a ring network that includes their respective servers and an aggregation switch. The routing capability provided via the network hardware components function within the network to advertise routes to each other and provide the aggregation switch access to pathways that can be used to deliver data packets to their destination without using other switches, e.g., a top-of-rack switch.