Packet Metadata Data Placement Across Heterogeneous CXL Devices

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Determining optimal data placement strategies across heterogeneous computing systems with new interconnect protocols like CXL, Gen-Z, and Slingshot to reduce latency and energy consumption is challenging for memory-intensive applications.

Innovation Solution

A system with a traffic analyzer that determines data placement based on packet metadata, using DMA engines to move data across devices and updating virtual-to-physical translations, with programmable event lists to manage data relocation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If data placement strategies are implemented across heterogeneous computing systems with new interconnect protocols, then latency and energy consumption are reduced, but system complexity increases

Engineering Contradiction:
ImprovelatencyVSAvoidsystem complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent introduces a traffic analyzer as an intermediary component that monitors packet metadata and triggers data movement events. This mediator sits between the heterogeneous interconnect protocols and the data placement logic, translating various protocol types into unified placement decisions without requiring complex integration of each protocol directly into the placement strategy

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system segments data placement functionality into independent components: packet metadata extraction, event triggering logic, and data movement execution. This segmentation allows each component to handle specific protocol types independently, reducing overall system complexity while maintaining low latency through specialized optimization of each segment

Inventive Principle:
Principle #1Segmentation

2Use of energy by moving object

If data placement optimization is implemented for memory-intensive applications, then energy consumption is reduced, but device complexity increases

Engineering Contradiction:
Improveenergy consumptionVSAvoiddevice complexity
Core Design Contradiction:
Use of energy by moving objectVSDevice complexity

Solution Approach 1:

The traffic analyzer autonomously monitors packet metadata and automatically triggers data movement events without requiring external control or complex management software. The system serves itself by using the existing packet metadata to make placement decisions, eliminating the need for additional energy-intensive monitoring and control mechanisms

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent changes the operational parameters of data movement by using packet metadata fields as triggers for placement decisions. This parameter-based approach allows the system to adapt to different application workloads and protocol types without increasing device complexity, as the same infrastructure handles varied scenarios through parameter variation

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4264905B1Data placement with packet metadata
Publication Date: 2025.08.13 ADVANCED MICRO DEVICES INC
  • EP4264905B1 patent drawingFigure 1
  • EP4264905B1 patent drawingFigure 2
  • EP4264905B1 patent drawingFigure 3

AI summary

Systems, apparatuses, and methods for determining data placement based on packet metadata are disclosed. A system includes a traffic analyzer that determines data placement across connected devices based on observed values of the metadata fields in actively exchanged packets across a plurality of protocol types. In one implementation, the protocol that is supported by the system is the compute express link (CXL) protocol. The traffic analyzer performs various actions in response to events observed in a packet stream that match items from a pre-configured list. Data movement is handled underneath the software applications by changing the virtual-to-physical address translation once the data movement is completed. After the data movement is finished, threads will pull in the new host physical address into their translation lookaside buffers (TLBs) via a page table walker or via an address translation service (ATS) request.