Virtual Machine Scaling via Pre-created Clusters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current virtual machine scaling in network function virtualization (NFV) systems faces high failure probabilities due to complex resource preemption and creation processes, leading to service disruptions and inefficient resource utilization.
Innovation Solution
A virtual machine scaling method that creates a virtual machine cluster for a VNF node, allowing for power-on or power-off operations with a single API invocation, reducing the need for re-application at the infrastructure layer and minimizing API invocation failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines are added or deleted for scaling purposes in NFV systems, then service pressure is relieved or resource utilization is improved, but the probability of scaling failure increases due to complex resource preemption and creation processes
Solution Approach 1:
The patent applies preliminary action by pre-creating a pool of virtual machines in advance before scaling operations are needed. These pre-created VMs are stored in a suspended state and can be quickly activated when scaling is required, eliminating the need for complex resource preemption and creation during actual scaling operations. This resolves the contradiction by improving both scaling efficiency (faster scaling) and reliability (higher success probability) through advance preparation.
2Ease of manufacture
If multiple API invocations are made during virtual machine creation and deletion, then resource allocation and release are achieved, but the probability of operation failure increases
Solution Approach 1:
The patent reduces API invocations during scaling operations by having already allocated resources (virtual machines) in advance. When scaling is needed, instead of making multiple API calls to create VMs and allocate resources, the system simply activates pre-created VMs from the suspended pool, requiring minimal API invocations. This maintains full resource management capability while significantly improving operation success probability by reducing failure points.
3Adaptability or versatility
If virtual machines are created on-demand during scaling operations, then resource utilization is optimized, but service disruption occurs due to scaling failures
Solution Approach 1:
The patent maintains service continuity by having virtual machines pre-created and ready in a suspended state before scaling operations. When scaling is required, these pre-prepared VMs can be rapidly activated without service disruption. This approach preserves resource flexibility (VMs can still be added or removed based on demand) while eliminating service disruptions caused by scaling failures, as the VMs are already created and merely need activation.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Embodiments of the present invention relate to the field of virtualization technologies, and provide a virtual machine scaling method and a virtual management device, to avoid a case in which a VM resource fails to be preempted to create a virtual machine for a VNF node, reduce a probability of a VNF node scaling failure, and avoid impact of VNF node scaling on a service of an NFV system to some extent. The method includes: creating, by a VNFM, a target VNF node, and creating a virtual machine cluster for the target VNF node; receiving, by the VNFM, a virtual machine scaling request, where the request carries an identifier of the target VNF node, information about a target virtual machine, and a virtual machine status identifier; determining, by the VNFM, the virtual machine cluster of the target VNF node, and determining the target virtual machine in the cluster; and setting the target virtual machine to a power-on state if the virtual machine status identifier is a power-on identifier, so as to scale out the target VNF node, or setting the target virtual machine to a power-off state if the virtual machine status identifier is a power-off identifier, so as to scale in the target VNF node.