Real-Time VNF Orchestration via CPU Core Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current orchestrators for cloud environments do not effectively manage real-time virtual network functions (VNFs), leading to degraded performance due to shared CPU resources and lack of real-time performance monitoring, which is critical for meeting latency requirements in mobile RAN processing.
Innovation Solution
A method and system for real-time virtual network function orchestration that involves reporting VNF performance, updating and isolating CPU cores, moving IRQs and system tasks, and managing indeterminism sources to ensure real-time performance, using a new orchestrator that prioritizes CPU resource allocation and monitoring based on VNF performance metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional orchestrators allocate CPU resources to multiple VNFs on shared cores, then resource utilization improves, but real-time performance degrades due to latency from I/O interrupts and context switching
Solution Approach 1:
The patent segments CPU cores into two distinct groups: real-time CPU cores dedicated to RT VNFs and non-real-time CPU cores for I/O interrupt threads. This segmentation eliminates the mixing of real-time and non-real-time tasks on the same core, ensuring that I/O interrupts do not interfere with real-time VNF execution and causing latency. The segmentation maintains high resource utilization by keeping both CPU groups actively engaged in their respective functions.
Solution Approach 2:
The patent extracts I/O interrupt threads from real-time CPU cores and relocates them to non-real-time CPU cores. This extraction removes the harmful factor (I/O interrupts) from the real-time processing environment, preventing them from degrading real-time performance. The I/O interrupt threads continue to function normally on their dedicated non-real-time cores, while real-time VNFs execute without interruption on isolated real-time cores.
2Adaptability or versatility
If multiple VNFs share CPU cores in cloud environments, then consolidation efficiency improves, but latency increases making it impossible to meet strict real-time requirements
Solution Approach 1:
The patent segments the CPU architecture into real-time and non-real-time domains, allowing VNFs to be consolidated on real-time CPU cores while I/O operations occur on separate non-real-time cores. This segmentation enables efficient consolidation of multiple RT VNFs without the latency penalty of shared I/O handling, as each RT VNF runs on dedicated real-time cores that do not contend with I/O interrupt processing.
Solution Approach 2:
The patent introduces non-real-time CPU cores as an intermediary layer that handles all I/O interrupt processing. This intermediary absorbs the time-consuming I/O operations, preventing them from directly impacting real-time VNF execution. The mediator approach allows the system to maintain high consolidation efficiency while isolating real-time processes from latency-inducing I/O activities.
3Device complexity
If traditional orchestrators manage VNF deployment without real-time monitoring, then system complexity remains low, but real-time performance constraints cannot be guaranteed
Solution Approach 1:
The patent implements a feedback mechanism where the orchestrator continuously monitors real-time performance metrics of VNFs and dynamically adjusts resource allocation. The orchestrator collects performance data from VNFs, compares it against real-time constraints, and automatically migrates VNFs between real-time and non-real-time CPU cores or adjusts their resource allocation to ensure constraint satisfaction. This closed-loop feedback system maintains reliability without requiring complex manual intervention.
Solution Approach 2:
The patent enables VNFs to self-report their real-time performance status and resource requirements to the orchestrator. VNFs autonomously monitor their own latency and throughput metrics, then communicate this information back to the orchestrator for decision-making. This self-service approach distributes the monitoring burden to individual VNFs, reducing the complexity burden on the central orchestrator while maintaining comprehensive real-time oversight.
Data Source
AI summary
Method, apparatus and system for real-time virtual network function orchestration in Real-time Cloud Infrastructure. The method comprises the step of updating and reporting CPU core RT performance; sending CPU and NUMA information; moving the IRQs from RT CPUset to Non-RT CPUset based on the CPU pool information response, and reporting the completion of IRQ clearance; moving the system tasks and OS background processes to Non-RT CPUset based on the CPU pool information response, and reporting the completion of system tasks clearance; clearing the known indeterminism sources based on the CPU pool information response, and reporting the completion of indeterminism sources clearance. The implementation of the method and apparatus improves that with support of NFV, edge cloud can speed new service deployment and achieve resource sharing among different services which allows operators to provision fewer resources.