Virtual Channel Data Transfer Between Peripheral Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Peripheral devices face inefficiencies in data transfers due to high host memory bandwidth consumption and processor latency, as they rely on host memory for data transfers, leading to conflicting demands for memory usage and processor response time.
Innovation Solution
A virtual channel system that presents a larger logical address space than the physical capacity of a memory buffer, allowing data transfers between peripheral devices without acknowledgments, managed by a transfer control module to control throughput and prevent buffer overflow.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If host memory is used for data transfers between peripheral devices, then data transfer capability is provided, but host memory bandwidth consumption increases and processor latency increases
Solution Approach 1:
The patent extracts the data transfer function from the host system by implementing direct peripheral-to-peripheral data paths that bypass host memory. The switching fabric enables peripheral devices to exchange data directly without involving host memory bandwidth, thereby resolving the contradiction between providing data transfer capability and reducing host memory bandwidth consumption.
Solution Approach 2:
The patent introduces a switching fabric as an intermediary component that mediates data transfers between peripheral devices. This switching fabric acts as a dedicated data path mediator that enables direct peripheral communication without requiring host memory involvement, thus eliminating the bandwidth consumption issue while maintaining transfer capability.
2Productivity
If host memory portion is increased to reduce processor latency effect, then processor usage is minimized, but additional latency is introduced while waiting for larger memory reads and writes
Solution Approach 1:
The patent removes data transfers from the host memory subsystem entirely by creating direct peripheral communication paths through the switching fabric. This extraction eliminates both the processor latency overhead and the memory access latency that would result from using larger host memory portions, as data never enters host memory in this architecture.
Solution Approach 2:
The patent creates a parallel data transfer pathway that copies the essential data movement function away from the host memory system. The switching fabric implements a separate, dedicated data path that replicates the data transfer capability without requiring host memory involvement, thus avoiding both processor and memory access latency.
3Loss of energy
If host memory portion is minimized to reduce memory usage, then memory bandwidth is saved, but processor response time latency increases and processor burden increases
Solution Approach 1:
The patent extracts the data transfer function completely from the host system, eliminating the need for host memory involvement in peripheral communications. This extraction simultaneously achieves minimal host memory bandwidth consumption and eliminates processor response time latency by enabling autonomous peripheral-to-peripheral data paths through the switching fabric.
4Productivity
If peripheral devices transfer data without acknowledgments, then data transfer efficiency is improved, but buffer overflow risk increases
Solution Approach 1:
The patent introduces the switching fabric as an intermediary that implements flow control mechanisms between peripheral devices. This mediator monitors buffer states and regulates data flow to prevent overflow while maintaining high transfer efficiency, as the switching fabric can manage multiple data streams and implement backpressure control without requiring acknowledgment protocols at the peripheral level.
Data Source
AI summary
Apparatuses, systems, and methods are disclosed for a virtual channel for data transfers between devices. A method includes presenting an address space for a memory buffer. The address space may be larger than a physical capacity of the memory buffer. A method includes controlling, from a peripheral device, a rate at which a data source transfers data to a memory buffer using a presented address space based on a rate at which a data target transfers data from the memory buffer using the presented address space so that an amount of data stored in the memory buffer remains at or below a physical capacity of the memory buffer.