Policy Synchronization via Unique ID Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In software-defined networks (SDNs), when a controller loses connectivity with the network management system (NMS), it leads to synchronization issues in the control plane, causing difficulties in predicting network behavior and requiring critical policy changes to be halted, which can result in urgent policy delays and potential reconnection issues.
Innovation Solution
Assigning an identifier to a policy and using it for synchronization within the control plane, where the NMS associates a unique ID with policies, exchanges these IDs with controllers, and manages connections to ensure only synchronized controllers receive updates, preventing delays and maintaining connections between controllers and routers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the NMS blocks policy push to all controllers until connectivity is restored, then policy synchronization is maintained, but critical policy changes are delayed and productivity decreases
Solution Approach 1:
The system segments the controller pool into synchronized and non-synchronized groups based on connectivity status. The NMS identifies individual controllers that have lost connectivity and handles them separately, while continuing to push policies to controllers that remain connected. This segmentation allows policy updates to proceed for available controllers without waiting for disconnected ones to reconnect.
Solution Approach 2:
The NMS performs partial policy push actions by sending policies to only those controllers that are currently reachable, rather than blocking the entire operation. This partial action ensures that policy updates are applied to available controllers immediately, improving productivity while maintaining reliability through selective synchronization.
2Reliability
If the NMS removes the disconnected controller from the valid controllers list, then policy synchronization is maintained, but the controller loses connection with routers and requires reestablishment
Solution Approach 1:
The NMS performs preliminary identification of disconnected controllers through heartbeat mechanisms before they can cause synchronization issues. By detecting connectivity loss early and handling these controllers separately in the valid list, the system prevents the need for complete connection reestablishment, reducing time loss while maintaining policy synchronization.
Solution Approach 2:
The NMS acts as an intermediary that manages the valid controllers list dynamically. It mediates between disconnected controllers and the policy push mechanism by identifying which controllers should temporarily be excluded from the valid list, allowing policy synchronization to proceed without forcing complete connection reestablishment and minimizing disruption time.
3Reliability
If a signaling mechanism is established between NMS and management software, then controller connectivity status is tracked, but system complexity increases
Solution Approach 1:
Controllers perform self-service by periodically sending heartbeat signals to the NMS to indicate their own connectivity status. This self-reported mechanism allows the NMS to automatically identify disconnected controllers and adjust the valid controllers list accordingly, providing reliable status tracking without requiring complex external signaling mechanisms between NMS and management software.
Solution Approach 2:
The system discards disconnected controllers from the valid controllers list temporarily and recovers them automatically when they reconnect. This simple discard-and-recover approach provides reliable status tracking through basic list management operations rather than complex signaling mechanisms, reducing system complexity while maintaining reliability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and system for synchronizing policy in a control plane are provided. The method includes associating, by a network management system (NMS), a first identifier (ID) with a first policy of a network, wherein the first ID uniquely identifies the first policy. The method also includes pushing, by the NMS, the first policy to the control plane including a plurality of controllers. Further, the method includes exchanging, by the NMS, the first ID with the plurality of controllers. Moreover, the method includes determining, by the NMS, at least one controller from which confirmation of the first ID is not received as a stale controller.