Hybrid SDN Architecture for Scalable Flow Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Software Defined Networking (SDN) with OpenFlow faces scalability issues due to the need for centralized control of millions of flows in large-scale data center networks, leading to high computational loads and costs for building scalable controllers.
Innovation Solution
A hybrid control architecture is introduced, separating network control into two tiers: macro-flows managed by SDN controllers and micro-flows managed by SDN network Edge devices in a distributed fashion, allowing Edge devices to learn flow information and compute paths locally without relying on southbound APIs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If centralized control is used for all flows in SDN, then flow management is simplified, but scalability deteriorates due to high computational load on controllers
Solution Approach 1:
The patent segments flow management into two tiers: macro-flows managed by centralized SDN controllers and micro-flows managed by distributed Edge devices. This segmentation allows the controller to focus on high-level network coordination while Edge devices handle granular flow details locally, resolving the scalability issue without sacrificing management simplicity.
Solution Approach 2:
The patent extracts micro-flow management responsibilities from the centralized controller and assigns them to Edge devices. By taking out the computational burden of micro-flow processing from the controller, the system maintains centralized oversight for macro-flows while distributing the heavy lifting to Edge devices, thereby improving controller scalability.
2Extent of automation
If millions of flows are managed by a single controller, then centralized control is maintained, but device complexity increases due to high processing requirements
Solution Approach 1:
The patent segments the control function by separating macro-flow control (centralized) from micro-flow control (distributed). This segmentation reduces the processing complexity on controllers by limiting them to macro-flow management only, while Edge devices handle micro-flow processing locally.
Solution Approach 2:
The patent introduces Edge devices as intermediary components between the centralized controller and end flows. These intermediaries absorb the complexity of micro-flow processing, allowing the controller to maintain centralized automation for macro-flows without bearing the full computational burden of all flows.
3Adaptability or versatility
If all flow information is collected by controllers, then centralized decision making is enabled, but network latency increases due to information gathering and processing delays
Solution Approach 1:
The patent extracts flow information collection and processing capabilities from the controller and places them at Edge devices. This extraction enables local flow setup decisions to be made immediately without waiting for controller processing, significantly reducing flow setup time while the controller retains centralized decision-making capability for macro-flows.
Solution Approach 2:
The patent enables Edge devices to perform preliminary flow processing and decision-making actions locally before controller involvement is needed. By preparing and executing micro-flow decisions at the Edge, the system eliminates delays associated with information gathering and processing at the controller level.
Data Source
AI summary
A segment based switching architecture with hybrid control providing flow control in software defined networking (SDN). An SDN controller controls top tier macro-flows and virtual segment backbone connections in a network, and an SDN network edge device controls and manages micro-flows that are attached to the SDN network edge device. The SDN network edge device controls and manages the micro-flows locally without using a southbound API. The SDN network edge device learns flow information such as source information and incoming port information for all unknown micro-flows belonging to defined segments, and builds a virtual segment topology database for each segment. The SDN network edge device employs flow computation algorithms and micro-flow management in a distributed fashion. The virtual segment backbone connections are naturally exposed to the SDN controller through provisioning.