Host Network Accelerators for Drop-Free Data Center Switch Fabric
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face challenges with proprietary switch fabrics that are costly and prone to single points of failure, while off-the-shelf packet-based switching components can result in lossy and nondeterministic behavior, and existing traffic management mechanisms are inadequate for tunnel networks like VPNs and IPv6 over IPv4.
Innovation Solution
The implementation of host network accelerators (HNAs) with integrated virtual routers and overlay forwarding technologies, using low-cost, industry-standard IP over Ethernet (IPoE) protocols, provides a scalable and drop-free data center switch fabric by embedding flow control, scheduling, and Quality of Service (QoS) features, enabling seamless packet transport across the switch fabric.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If proprietary switch fabric is used, then high performance is achieved, but cost increases and single point of failure risk arises
Solution Approach 1:
The system segments the switch fabric functionality by implementing virtual switch instances distributed across multiple host network accelerators instead of using a single proprietary switch fabric. Each HNA contains a virtual switch instance that can independently forward packets, creating multiple failure domains and eliminating the single point of failure while maintaining high performance through parallel processing across multiple devices.
2Ease of manufacture
If off-the-shelf packet-based switching components are used, then cost is reduced, but lossy and nondeterministic behavior occurs
Solution Approach 1:
The system implements a feedback mechanism where the virtual switch monitors buffer occupancy and flow control status, then dynamically adjusts packet forwarding decisions. The virtual switch tracks the amount of data in buffers and uses this information to make informed forwarding decisions, preventing buffer overflow and packet loss while maintaining deterministic behavior. This feedback loop enables reliable packet delivery using commodity hardware that would otherwise exhibit lossy behavior.
3Adaptability or versatility
If overlay forwarding technologies are implemented, then scalability is improved, but device complexity increases
Solution Approach 1:
The system introduces an intermediary layer in the form of encapsulation headers that carry virtual network identification information. Instead of modifying the core switching logic to handle complex virtual network routing directly, the overlay forwarding mechanism uses standardized encapsulation protocols (such as VXLAN or GRE) that add metadata to packets. This intermediary layer allows the virtual switch to identify and forward packets across different virtual networks using simple lookup tables, achieving scalability without proportionally increasing device complexity.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A high-performance, scalable and drop-free data center switch fabric and infrastructure is described. The data center switch fabric may leverage low cost, off-the-shelf packet-based switching components (e.g., IP over Ethernet (IPoE)) and overlay forwarding technologies rather than proprietary switch fabric. In one example, host network accelerators (HNAs) are positioned between servers (e.g., virtual machines or dedicated servers) of the data center and an IPoE core network that provides point-to-point connectivity between the servers. The HNAs are hardware devices that embed virtual routers on one or more integrated circuits, where the virtual router are configured to extend the one or more virtual networks to the virtual machines and to seamlessly transport packets over the switch fabric using an overlay network. In other words, the HNAs provide hardware-based, seamless access interfaces to overlay technologies used for communicating packet flows through the core switching network of the data center.