Unified Resource Allocation for Virtual Network Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current approaches to allocating CPU cycles and memory in virtualized systems for virtual network functions (VNFs) are independent, leading to inefficient resource utilization and potential degradation in throughput and service provisioning due to the lack of coordination between these resources.
Innovation Solution
The proposed solution involves categorizing loads into committed and peak categories to allocate CPU cycles and memory based on packet loss permissions, using expressions to calculate the effective number of vCPUs and memory sizes that consider both committed and peak loads, and employing fluid-flow models to optimize resource allocation for statistical multiplexing and service level agreements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If CPU cycles and memory are allocated independently for VNFs, then resource allocation simplicity is maintained, but resource utilization efficiency deteriorates
Solution Approach 1:
The patent merges CPU cycle allocation and memory allocation into a unified resource allocation process. By considering both resources simultaneously and calculating their requirements together based on traffic flow characteristics, the system achieves coordinated allocation that improves resource utilization efficiency while maintaining manageable complexity through integrated management.
2Reliability
If resource allocation is optimized for peak load, then service reliability during high traffic is improved, but resource waste during low traffic increases
Solution Approach 1:
The patent implements dynamic resource allocation that adapts to varying traffic conditions. By categorizing traffic flows into committed and peak load components, the system allocates resources dynamically - ensuring sufficient capacity during peak periods for reliability while allowing resource release during low traffic periods to minimize waste, rather than maintaining static over-provisioned resources.
3Productivity
If statistical multiplexing is employed, then resource sharing efficiency is improved, but packet loss risk increases
Solution Approach 1:
The patent segments traffic flow requirements into two distinct categories: committed load that guarantees minimum service levels and prevents packet loss, and peak load that allows statistical multiplexing with acceptable packet loss. This segmentation enables the system to share resources efficiently through statistical multiplexing for peak demands while maintaining reliable service for committed requirements through guaranteed resource reservations.
Data Source
AI summary
A method, a device, and a non-transitory storage medium are described in which a resource allocation service is provided in relation to a virtual device. The resource allocation service calculates an allocation of a shared processor and a shared memory in support of the virtual device based on whether packet loss is permitted or not. The calculation of the processor allocated to the virtual device may be based on buffer memory allocation. Alternatively, the calculation of the processor allocated to the virtual device may be based on a packet loss ratio and a buffer memory allocation.