Transaction ID Encoding for Multiprocessor Bandwidth Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In multiprocessor computer systems, the increasing size of communications required for coherency protocols leads to higher network bandwidth needs and increased costs, necessitating a reduction in the amount of information sent between components.

Innovation Solution

Implementing multiple classes of transaction IDs, where the sending device encodes only a portion of the ID in initiation packets and the receiving device regenerates the full ID, allowing for efficient communication by classifying transactions based on packet characteristics and using a subset ID for identification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the full transaction ID is encoded in every packet, then transaction identification accuracy is improved, but packet size increases and network bandwidth requirements increase

Engineering Contradiction:
Improvetransaction identification accuracyVSAvoidnetwork bandwidth
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The transaction ID is divided into two parts: a subset ID that is encoded in initiation packets, and a packet characteristic ID that is derived from packet characteristics. This segmentation allows the sending device to transmit only the essential subset ID in the initial packet, reducing packet size, while the receiving device reconstructs the full transaction ID by combining the subset ID with the derived packet characteristic ID.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The packet characteristic ID acts as an intermediary element that bridges the subset ID and the full transaction ID. Instead of transmitting the complete transaction ID directly, the system uses packet characteristics (such as packet type, direction, or other identifiable features) as an intermediary to generate the packet characteristic ID, which when combined with the subset ID, reconstructs the full transaction ID at the receiving device.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If multiple classes of transaction IDs are implemented, then communication efficiency is improved, but device complexity increases

Engineering Contradiction:
Improvecommunication efficiencyVSAvoidtransaction ID management complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system dynamically determines which packets require full transaction ID encoding based on their characteristics. Initiation packets transmit only the subset ID, while responsive packets encode the full transaction ID. This dynamic approach allows the system to optimize communication efficiency for different packet types without requiring complex static configuration, as the encoding behavior is automatically determined by packet characteristics.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentUS7571252B2Computer system with multiple classes of transaction IDs
Publication Date: 2009.08.04 ORACLE AMERICAN INC
  • US7571252B2 patent drawing
  • US7571252B2 patent drawing
  • US7571252B2 patent drawing

AI summary

A computer system may include a sending device, a receiving device, and a network coupling the devices. The sending device may be configured to send a packet on the network in order to initiate a transaction. The sending device is configured to only encode a portion of a transaction ID identifying the transaction in the packet. The receiving device is configured to receive the packet from the network and to send a responsive packet to the sending device as part of the transaction. The receiving device is configured to encode all of the transaction ID in the responsive packet. The receiving device may generate the portion of the transaction ID not encoded in the packet by the sending device in response to the packet having a particular characteristic.