Hardware Containerization Framework for Wireless SoC Resource Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wireless network architectures are limited in managing hardware resources efficiently, particularly in provisioning and metering non-CPU processing elements for real-time Quality of Service (QoS) guarantees, as they primarily focus on CPU-based resources, neglecting the need for custom workloads like AI and RAN.
Innovation Solution
A hardware containerization framework is introduced that enables the management of hardware resources by selecting and configuring hardware units based on capability information, allowing for the allocation, partitioning, and isolation of resources like GPUs and FPGAs, and providing mechanisms for metering and optimizing their usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional CPU-based resource management is used, then system compatibility and ease of operation are maintained, but resource utilization efficiency and power efficiency deteriorate due to inability to properly manage heterogeneous hardware resources
Solution Approach 1:
The patent segments hardware resources into distinct containerizable units (CPU cores, GPUs, FPGAs, NPU, DSP) that can be independently managed and allocated. Each hardware unit type is treated as a separate resource pool that can be partitioned and assigned to different workloads, enabling efficient utilization of heterogeneous resources without requiring a monolithic management approach.
Solution Approach 2:
The patent creates a universal containerization framework that can manage multiple types of heterogeneous hardware resources (CPU, GPU, FPGA, NPU, DSP) through a common interface and management mechanism. This multi-functional approach allows the same containerization infrastructure to handle diverse workloads including AI processing, RAN functions, and general computing tasks, improving resource utilization across different hardware types.
2Power
If heterogeneous multi-core processors are used to improve power and performance, then processing capability is enhanced, but device complexity and difficulty of resource allocation increase
Solution Approach 1:
The patent implements dynamic resource allocation where container configurations can be modified at runtime based on workload requirements. The system can dynamically assign or reassign hardware resources (CPU cores, GPUs, FPGAs) to different containers as needed, allowing flexible adaptation to changing processing demands and optimizing power efficiency without being locked into static allocations.
Solution Approach 2:
The containerization framework acts as an intermediary layer between the heterogeneous hardware resources and the applications. This intermediate abstraction layer simplifies resource management by providing standardized interfaces and allocation mechanisms, reducing the complexity of directly managing multi-core processors and various accelerators while maintaining the ability to optimize power and performance.
3Use of energy by moving object
If hardware containerization is implemented to improve resource allocation, then resource utilization and power efficiency are enhanced, but device complexity and implementation difficulty increase
Solution Approach 1:
The patent performs preliminary configuration of hardware resources into containerizable units before runtime execution. Hardware resources are pre-identified, categorized, and made available as configurable units that can be allocated to containers. This preliminary preparation simplifies runtime resource management and reduces implementation complexity by establishing a structured framework in advance.
Solution Approach 2:
The patent creates abstract container representations that copy or mirror the essential resource allocation characteristics without requiring physical duplication of hardware. Container configurations capture the logical structure and resource requirements, enabling virtual resource management that simplifies implementation while maintaining the ability to optimize actual hardware utilization and power efficiency.
4Measurement precision
If capability information collection from hardware units is implemented, then resource allocation accuracy is improved, but measurement and detection complexity increases
Solution Approach 1:
The patent implements a universal capability information collection mechanism that works across different hardware unit types (CPU, GPU, FPGA, NPU, DSP) through standardized interfaces. This multi-functional approach allows the system to gather capability information from diverse hardware sources using a common method, improving measurement precision without proportionally increasing detection complexity through hardware-specific implementations.
Data Source
AI summary
System and method for managing hardware containerization in a wireless network architecture, are described. In one aspect, capability information of a plurality of hardware units is transmitted to a host processor to select a set of hardware units in a system-on-chip (SoC). Configuration information that includes the selection of one or more hardware allocation parameters and one or more types of statistics for one or more processing flows is received by the selected set of hardware units. A hardware container is configured to enable SoC resource allocation related to a control group and a namespace of SoC resources via a plurality of hardware modules in the SoC. The hardware container is managed based on periodic collection and analysis of the statistical data via a plurality of instrumentation modules in the SoC, and the impact on the one or more processing flows is tracked at the host processor.