Disaggregated Memory Packet Scheduling With Distributed Time Slots

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current Time Division Multiple Access (TDMA) approaches for memory systems in large-scale data centers suffer from scalability limitations, delayed adaptation to changing data flows, heavy operational overhead, and hardware requirements, making them unsuitable for dynamic network demands in cloud systems.

Innovation Solution

A distributed time scheduling method where processing nodes manage their own scheduling, reducing the need for centralized synchronization and enabling faster adaptation to network demands by using localized traffic load and packet reception capacity information to assign time slots to transmission queues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If centralized TDMA scheduling is used throughout the network, then performance guarantee for different applications is improved, but scalability is restricted and operational overhead increases

Engineering Contradiction:

Improveperformance guaranteeVSAvoidoperational overhead

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the centralized scheduling function into distributed scheduling units at each network device. Each device independently performs TDMA scheduling for its local traffic flows, eliminating the need for a central controller and reducing operational overhead while maintaining performance guarantees through localized resource allocation

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Each network device autonomously performs scheduling decisions based on local traffic conditions and pre-configured policies. The devices self-manage their time slot allocations and adapt to changing traffic patterns without requiring centralized coordination, reducing complexity while preserving reliability

Inventive Principle:

Principle #25Self-service

2Reliability

If centralized TDMA scheduling is used throughout the network, then performance guarantee for different applications is improved, but scalability is restricted

Engineering Contradiction:

Improveperformance guaranteeVSAvoidscalability

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The scheduling function is divided into independent segments at each network device, allowing the system to scale horizontally by simply adding more autonomous devices. Each segment operates independently, enabling the network to grow without requiring reconfiguration of a central controller

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Each network device is equipped with universal scheduling capabilities that allow it to function as both a data forwarding node and a scheduling decision-maker. This multi-functionality enables scalable deployment where any device can independently provide TDMA scheduling services

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Ease of operation

If centralized TDMA scheduling is used throughout the network, then fair network bandwidth allocation is improved, but adaptation to changing data flows is delayed

Engineering Contradiction:

Improvefair bandwidth allocationVSAvoidadaptation speed

Core Design Contradiction:

Ease of operationVSSpeed

Solution Approach 1:

Fair bandwidth allocation policies are pre-configured at each distributed scheduling unit before traffic flows begin. When traffic patterns change, each device immediately applies its local policies without waiting for centralized reconfiguration, achieving both fairness and rapid adaptation

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Each network device continuously monitors local traffic conditions and automatically adjusts time slot allocations based on observed flow patterns. This local feedback mechanism enables rapid adaptation to changing data flows while maintaining fair bandwidth distribution according to configured policies

Inventive Principle:

Principle #23Feedback

4Productivity

If centralized TDMA scheduling is used throughout the network, then network bandwidth allocation is improved, but hardware overhead and operational complexity increase

Engineering Contradiction:

Improvenetwork bandwidth allocationVSAvoidhardware overhead

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

Each network device autonomously manages its own bandwidth allocation using locally available resources and pre-configured policies. This self-service approach eliminates the need for dedicated centralized hardware controllers and reduces overall system complexity while maintaining efficient bandwidth utilization

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS12568052B2Data transmission scheduling for disaggregated memory systems

Publication Date: 2026.03.03 WESTERN DIGITAL TECHNOLOGIES INC

AI summary

A processing node executes one or more applications that are allocated memory of at least one shared memory of one or more memory nodes via a network. Packets including memory messages for memory nodes are enqueued into one or more transmission queues for sending the packets to the memory nodes. Each packet is enqueued into a transmission queue according to the application issuing the memory message included in the packet. One or more time slots of a predetermined time period are assigned to a transmission queue for sending at least a portion of the packets enqueued in the transmission queue within the assigned one or more time slots. At least a portion of the packets from the transmission queue are sent during the one or more assigned time slots. In one aspect, time slots are assigned based on scheduling information from a network scheduling device.