CPU Power State Pooling for Heterogeneous vCPU Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack effective methods for optimizing energy efficiency and power consumption in large-scale virtualized environments, particularly in NFV and vRAN, due to conflicting CPU power state requirements among heterogeneous virtualized workloads, which existing power management techniques fail to address at a global level.
Innovation Solution
Implement power state-based resource pooling and scheduling of virtual CPUs (vCPUs) on physical cores that match their power profile characteristics, using a power state controller to dynamically adjust CPU power states based on real-time utilization metrics, and integrating this with orchestration systems for optimal power management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If CPU power states are dynamically adjusted at the infrastructure level, then power consumption is reduced, but conflicts arise between physical processor P-state and desired P-state of vCPU scheduled on that physical core
Solution Approach 1:
The patent segments the physical CPU cores into multiple resource pools based on their power state characteristics (e.g., high-performance pools, energy-efficient pools). Each pool is associated with specific P-state ranges, allowing the system to match vCPUs with appropriate physical cores based on power requirements. This segmentation resolves the contradiction by preventing P-state conflicts through dedicated pool assignments rather than dynamic adjustments on shared cores.
Solution Approach 2:
The system performs preliminary classification of physical CPU cores into resource pools with defined P-state characteristics before workload scheduling. By pre-establishing these pools with known power state properties, the orchestrator can make informed scheduling decisions that avoid P-state conflicts from the outset, rather than attempting dynamic adjustments that may cause conflicts.
2Productivity
If multiple vCPUs share the same physical CPU cores on a time-sharing basis, then resource utilization is improved, but it becomes difficult to meet diverse power and performance requirements of different VNFCs
Solution Approach 1:
The patent divides the shared physical CPU resources into segmented resource pools, where each pool is characterized by specific power state ranges. This segmentation allows the system to maintain high resource utilization by sharing pools among multiple VNFCs while simultaneously meeting diverse power requirements by assigning VNFCs to appropriate pools based on their power profiles.
Solution Approach 2:
Different resource pools are assigned different local qualities in terms of power state characteristics. High-performance pools operate at higher P-states for performance-critical VNFCs, while energy-efficient pools operate at lower P-states for less demanding workloads. This local quality differentiation enables the system to accommodate diverse power and performance requirements within a shared infrastructure.
3Loss of energy
If power state management is implemented at the infrastructure level, then global power optimization is achieved, but coordination complexity between orchestration and scheduling increases
Solution Approach 1:
The patent introduces resource pools as an intermediary layer between the infrastructure's physical CPU cores and the orchestration system's vCPU scheduling. These pools abstract the complex power state management details, presenting a simplified interface to the orchestrator while enabling global power optimization through pool-based resource allocation and matching.
Data Source
AI summary
A method manages virtual network function components (VNFCs) in a compute infrastructure of a virtualized environment. The method includes performing, by an orchestrator system of the virtualized environment, power state-based resource pooling in the compute infrastructure based on central processing unit (CPU) power state management policies; and scheduling, by a scheduler, virtual CPUs (vCPUs) of the VNFCs on physical cores of appropriate resource pools whose CPU power states match with power profile characteristics of the vCPUs.