Flexible Resource Sharing in Network Processing Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data center architectures face challenges in efficiently managing data transfers and resource allocation due to limitations in traditional communication protocols, which struggle to keep pace with increasing network speeds and dynamic workloads, particularly in supporting latency-sensitive processing and memory-intensive applications.
Innovation Solution
The implementation of a Compute Express Link (CXL)-based communication system within data center clusters, enabling high-speed CPU-to-device and CPU-to-memory interconnects, maintaining memory coherency, and allowing for dynamic protocol multiplexing, resource sharing, and efficient data transmission across accelerators and memory devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional communication protocols are used in data center architectures, then device complexity and protocol overhead are reduced, but network speed and data transfer efficiency cannot keep pace with increasing requirements
Solution Approach 1:
The patent implements dynamic protocol selection and switching capabilities. The system can dynamically choose between different communication protocols (PCIe, CXL, InfiniBand, RoCE, RDMA) based on real-time workload requirements, enabling high-speed data transfer when needed while maintaining simplicity for常规 operations.
Solution Approach 2:
The network processing device is designed to support multiple communication protocols simultaneously, making it universally applicable to various workloads. This multi-protocol capability allows the same hardware to handle different data transfer requirements without requiring separate dedicated systems for each protocol type.
2Adaptability or versatility
If dynamic workload changes are accommodated, then adaptability and resource utilization are improved, but latency in resource allocation and protocol switching increases
Solution Approach 1:
The system performs preliminary actions by pre-configuring multiple communication paths and protocols before workload changes occur. The network processing device maintains ready-to-use connections and pre-established resource allocations, enabling rapid switching between workloads without requiring time-consuming setup procedures.
Solution Approach 2:
The patent implements feedback mechanisms that continuously monitor workload characteristics and resource usage patterns. Based on this feedback, the system automatically adjusts protocol selection and resource allocation in real-time, reducing latency by making informed decisions about optimal resource assignment without manual intervention.
3Reliability
If memory coherency is maintained across distributed systems, then data consistency and reliability are improved, but system complexity and synchronization overhead increase
Solution Approach 1:
The patent introduces coherency agents as intermediary components that manage memory coherency automatically. These agents act as mediators between different memory subsystems, handling the complex synchronization tasks and abstracting coherency management from the main system architecture, thereby maintaining data consistency without proportionally increasing overall system complexity.
4Adaptability or versatility
If resource sharing is enabled across multiple devices, then resource utilization and flexibility are improved, but access control complexity and potential conflicts increase
Solution Approach 1:
The patent segments resources into fine-grained units that can be independently allocated and managed. By dividing large resources into smaller, manageable portions, the system enables flexible sharing while simplifying access control through precise, unit-level permissions and allocation tracking.
Data Source
AI summary
A network processing device connects to one or more devices in a computing node and connects to one or more other network processing devices of other computing nodes. The network processing device identifies a policy for allowing devices in other computing nodes to access a particular resource of one of the devices in its computing node. The network processing device receives an access request to access the particular resource from another network processing device and sends a request to the device hosting the particular resource based on the access request and the policy.