Data Center Network Flat Addressing and Mesh Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data center architectures face issues such as resource fragmentation, poor server-to-server connectivity, and reliance on proprietary hardware that scales up rather than out, leading to bottlenecks and inefficiencies in handling increasing server demands.
Innovation Solution
A simple mesh-like architecture using commodity switches and modifications to the control plane for flexible traffic engineering, allowing for flat addressing and multipath routing, which disaggregates load balancing functionality across standard servers, enabling dynamic server pooling and optimized bandwidth between all servers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional architecture with IP routers and load balancers is used, then server connectivity is provided, but network bottlenecks and high costs occur
Solution Approach 1:
The patent segments the network into hierarchical levels (pod networks, rack networks, blade networks) with different addressing schemes for each level. This allows complex connectivity requirements to be broken down into manageable segments, eliminating the need for complex IP routing while maintaining full server-to-server connectivity.
Solution Approach 2:
The patent introduces a network address translation (NAT) mechanism as an intermediary that translates between flat addresses used by servers and the hierarchical physical addresses used by switches. This intermediary layer simplifies the network architecture by abstracting away the complexity of hierarchical addressing from server applications.
2Ease of operation
If proprietary hardware is used for load balancing and routing, then traffic management is achieved, but cost and scalability are worsened
Solution Approach 1:
The patent makes all servers universal participants in load balancing by enabling any server to act as a load balancer for any other server through the flat address translation mechanism. This eliminates the need for specialized proprietary load balancer hardware and allows dynamic, flexible traffic management across the entire server fleet.
Solution Approach 2:
The patent copies the load balancing functionality from specialized hardware to standard server hardware through software implementation. By translating flat addresses to physical addresses at the network switch level, the system replicates load balancing capabilities across all servers without requiring dedicated load balancer devices.
3Device complexity
If VLANs are used to segment networks, then control plane overhead is reduced, but resource fragmentation occurs
Solution Approach 1:
The patent creates an equipotential network environment where all servers have equal access to all resources through the flat addressing scheme. Every server can directly address and communicate with every other server without being constrained by VLAN boundaries, eliminating resource fragmentation while maintaining manageable control plane overhead through hierarchical switching.
4Ease of operation
If hierarchical network architecture is used, then network management is simplified, but bandwidth and connectivity are reduced
Solution Approach 1:
The patent applies local quality by implementing hierarchical switching with different addressing and routing characteristics at different levels of the network hierarchy. Each level (pod, rack, blade) has optimized local connectivity with direct flat address translation, while higher levels provide aggregation and routing. This ensures full bandwidth is available locally while maintaining simplified hierarchical management overall.
Data Source
AI summary
A method of networking a plurality of servers together within a data center is disclosed. The method includes the step of addressing a data packet for delivery to a destination server by providing the destination server address as a flat address. The method further includes the steps of obtaining routing information required to route the packet to the destination server. This routing information may be obtained from a directory service servicing the plurality of servers. Once the routing information is obtained, the data packet may be routed to the destination server according to the flat address of the destination server and routing information obtained from the directory service.