Peer-to-Peer PCIe Fabric Data Transfer via Management Processor
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As networked storage systems and remote computing systems increase in density, physical limitations such as space requirements for interconnects and climate control systems become constraints, and traditional bulk storage systems are limited in the number of devices per host, leading to issues with capacity, redundancy, and reliability, especially in networked, cloud, and enterprise environments.
Innovation Solution
The system employs a management processor to establish a communication arrangement between endpoint devices coupled to a communication fabric, redirecting data transfers without passing through a host processor, enabling peer-to-peer data transfers within the PCIe fabric, and allowing logical partitioning of components for efficient resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data transfers pass through host processor, then device compatibility and control are maintained, but transfer speed and resource utilization deteriorate
Solution Approach 1:
The communication fabric is segmented into virtual channels that can be independently configured and managed. Each virtual channel provides dedicated peer-to-peer communication paths, allowing data transfers to be routed directly between endpoint devices without involving the host processor, thereby improving transfer speed while maintaining overall system control through virtual channel management
Solution Approach 2:
A management processor acts as an intermediary to establish and configure communication arrangements between endpoint devices. The management processor sets up virtual channel connections and routing without participating in actual data transfers, enabling fast peer-to-peer communication while maintaining system-wide coordination and control
2Quantity of substance
If storage density increases, then capacity per unit space improves, but physical limitations such as interconnect space and climate control requirements worsen
Solution Approach 1:
The patent transitions from traditional host-mediated data paths to direct peer-to-peer communication paths between endpoint devices within the communication fabric. This dimensional change in data flow architecture eliminates the need for additional host processor resources and interconnect space, allowing storage density to increase without proportionally increasing physical space requirements for interconnects and climate control systems
3Quantity of substance
If number of devices per host increases, then storage capacity and redundancy improve, but host processor burden and system complexity worsen
Solution Approach 1:
The system segments communication paths into dedicated virtual channels for each endpoint device pair. This segmentation allows multiple devices to communicate independently through the fabric without requiring the host processor to manage each individual communication, enabling high device counts while maintaining manageable system complexity through automated virtual channel routing
Solution Approach 2:
Endpoint devices are empowered to initiate and complete peer-to-peer data transfers autonomously through configured virtual channels without requiring host processor intervention. This self-service capability allows numerous devices to be connected to the fabric while the host processor remains unburdened by individual transfer management, reducing system complexity despite high device density
Data Source
AI summary
Computing architectures, platforms, and systems are provided herein. In one example, system is provided. The system includes a management processor configured to initiate a communication arrangement between a first endpoint device coupled to a communication fabric and a second endpoint device coupled to the communication fabric. The communication arrangement is configured to redirect a transfer from the first endpoint device based on an address corresponding to an address range of the second endpoint device without passing the transfer through a host processor coupled to the communication fabric that executes an application initiating the transfer.