Receiver-Side Packet Ordering for Link-Rate Throughput

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current packet ordering mechanisms in networks face limitations in providing fine-grained ordering, leading to suboptimal performance due to round-trip delays and restrictive routing, which hinders the exploitation of available path diversity and results in reduced throughput and increased latency.

Innovation Solution

The proposed solution involves speculative connection setup, target-side ordering, fine-grained ordering, and ordering relaxations to enable link-rate performance by avoiding round-trip delays and cumulative dependencies, allowing for multi-packet transfers over unordered and unreliable networks while ensuring reliable data operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If transmitter-side ordering is used to ensure packet delivery order, then ordering reliability is improved, but throughput deteriorates due to serialization limiting transfer rate to one packet per RTT

Engineering Contradiction:
Improvepacket delivery orderVSAvoidthroughput
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent inverts the traditional transmitter-side ordering approach by implementing receiver-side ordering. Instead of the transmitter serializing packet transfers and waiting for acknowledgments before injecting the next packet, the receiver performs the ordering function. This allows the transmitter to inject packets at full link rate without waiting for RTT delays, while the receiver reorders incoming packets to ensure in-order delivery to the application layer.

Inventive Principle:
Principle #13The other way round (Inversion)

Solution Approach 2:

The patent segments the ordering function from the transmission function. The transmitter focuses on high-speed packet injection without ordering constraints, while the receiver handles the reordering task using sequence numbers and buffers. This segmentation allows independent optimization of throughput (at transmitter) and ordering reliability (at receiver).

Inventive Principle:
Principle #1Segmentation

2Reliability

If deterministic routing is used to provide ordering, then ordering reliability is improved, but adaptability deteriorates due to restriction to single path between source and target

Engineering Contradiction:
Improvepacket delivery orderVSAvoidpath diversity
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent inverts where ordering is enforced: instead of using deterministic routing to guarantee order, it allows any routing path and performs ordering at the receiver. This enables exploitation of path diversity and load balancing across multiple routes while maintaining ordering through sequence numbers and reordering buffers at the destination.

Inventive Principle:
Principle #13The other way round (Inversion)

Solution Approach 2:

The patent introduces sequence numbers as an intermediary mechanism that decouples routing from ordering. Packets can take any path through the network, but each packet is tagged with a sequence number that allows the receiver to reconstruct the original order regardless of arrival sequence. This intermediary enables both flexible routing and reliable ordering.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS11470394B2Scalable light-weight protocols for wire-speed packet ordering
Publication Date: 2022.10.11 NVIDIA CORP
  • US11470394B2 patent drawing
  • US11470394B2 patent drawing
  • US11470394B2 patent drawing

AI summary

A communication method between a source device and a target device utilizes speculative connection setup between the source device and the target device, target-device-side packet ordering, and fine-grained ordering to remove packet dependencies.