VNF Deployment Through Node-Group Segmentation for Fault Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In a containerization scenario, the increasing number of virtualized network functions (VNFs) sharing resources in a host aggregate leads to label stacking, increased management costs, and heightened fault impact due to the lack of efficient co-deployment strategies, affecting all VNFs on a faulty board.
Innovation Solution
Implementing a method that deploys VNFs by planning node labels within specific node groups, reducing management costs and minimizing fault impact by ensuring VNFs are scheduled and matched only within their designated groups, using a CaaS manager to manage node groups and resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple VNFs share resources in a host aggregate, then resource utilization improves, but management costs increase due to label stacking
Solution Approach 1:
The host aggregate is segmented into multiple node groups, where each node group is assigned to a specific VNF. This segmentation prevents label stacking by ensuring that each VNF only manages labels within its designated node group, rather than across the entire host aggregate. The CaaS manager enforces this segmentation by validating that VNFs can only schedule and match nodes within their assigned node groups.
2Adaptability or versatility
If VNFs are deployed on any board in the HA, then deployment flexibility improves, but fault impact increases when a board fails
Solution Approach 1:
The host aggregate is divided into multiple node groups, with each node group assigned to a specific VNF. This segmentation creates isolated deployment zones where VNFs can only schedule and match nodes within their own node group. This maintains deployment flexibility within bounds while preventing fault propagation across different VNFs, as a board failure only affects the specific node group and its associated VNF.
3Adaptability or versatility
If each VNF independently configures node labels, then VNF autonomy improves, but label stacking occurs increasing management overhead
Solution Approach 1:
The label management space is segmented by assigning each VNF a dedicated node group. Within this node group, each VNF can independently configure and manage its own labels without interfering with other VNFs. The CaaS manager enforces this segmentation by validating that VNFs only operate within their assigned node groups, thereby maintaining VNF autonomy while eliminating label stacking and reducing management overhead.
4Productivity
If VNFs share the same node pool, then resource sharing efficiency improves, but consequential impact between VNFs worsens
Solution Approach 1:
The node pool is segmented into multiple node groups, with each node group assigned to a specific VNF. This creates isolated resource domains where each VNF has dedicated access to its assigned nodes. The CaaS manager enforces this segmentation by validating that VNFs can only schedule and match nodes within their own node group, thereby preventing consequential impact between VNFs while maintaining resource sharing efficiency within each node group.
Data Source
AI summary
A method for deploying a virtualized network function (VNF) is provided. The method includes a first network element sends a first request message to a container as a service (CaaS) manager for requesting to deploy a first virtual deployment unit (VDU) of a first VNF, where the first request message comprises requirement information of the first VDU, the requirement information of the first VDU comprises requirement information that is of a node group and that is for deploying the first VDU, and the node group is a set of one or more nodes in one host aggregate (HA). The first network element receives, from the CaaS manager, a first request response message indicating that the first VDU is successfully deployed, the first VDU is deployed on a first node belonging to a first node group, and the first node group meets the requirement information of the node group.