DDR Memory Packet Processing with Descriptor Integration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Network devices face inefficiencies in packet processing due to separate storage and reading of packet entities (PKT) and packet descriptors (PD) in different memory types, leading to increased design costs and wiring pressure, as well as low reading efficiency.

Innovation Solution

Storing both packets and packet descriptors in the same DDR memory by dividing it into blocks with separate storage spaces, allowing for simultaneous storage and reading of packets and PDs, thereby optimizing memory utilization and reducing design costs and wiring complexity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If PD is stored in QDR SRAM and PKT is stored in DDR SDRAM separately, then the packet processing can be performed, but the design cost increases and wiring pressure increases

Engineering Contradiction:
Improvepacket processing capabilityVSAvoiddesign cost and wiring pressure
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the storage of PD and PKT into a single DDR memory by allocating different memory regions: the first memory region stores PKT while the second memory region stores PD. This consolidation eliminates the need for separate QDR SRAM for PD storage, thereby reducing design cost and wiring pressure while maintaining packet processing capability.

Inventive Principle:
Principle #5Merging (Combining)

2Device complexity

If PD and PKT are read separately in two steps, then the storage structure is simple, but the reading efficiency is low

Engineering Contradiction:
Improvestorage structure simplicityVSAvoidreading efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent enables continuous reading of PD and PKT by storing them in the same DDR memory with adjacent or logically connected memory addresses. The network device can continuously access both PD and PKT without switching between different memory types, eliminating idle time and improving reading efficiency while keeping the storage structure relatively simple.

Inventive Principle:
Principle #20Continuity of useful action

3Speed

If PD is stored in QDR SRAM, then the access speed to PD is fast, but the design cost increases

Engineering Contradiction:
ImprovePD access speedVSAvoiddesign cost
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The patent replaces the expensive QDR SRAM with DDR memory for storing PD. Although DDR memory has lower raw access speed compared to QDR SRAM, the overall system performance is improved by eliminating the need for expensive QDR SRAM hardware while maintaining adequate access speed through optimized memory management and continuous access patterns.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Data Source

PatentUS11425057B2Packet processing
Publication Date: 2022.08.23 NEW H3C TECH CO LTD
  • US11425057B2 patent drawing
  • US11425057B2 patent drawing
  • US11425057B2 patent drawing

AI summary

A memory of a network device is divided into first blocks, each first block being divided into second blocks, and each second block including a first storage space and second storage space. When a packet is stored, second blocks occupied by the packet are determined based on a packet length and a first storage space length, and the packet is stored into a first storage space of each of the second blocks. For each of the second blocks, a PD corresponding to the second block is generated, and stored into a second storage space of the second block. When a packet is read, the second blocks to be read are determined based on a start address of the packet. A packet fragment is read from a first storage space of the second blocks to be read, and the read packet segments are composed into the second packet to be sent.