Coherent Accelerator Firmware Update Without System Reboot

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for updating accelerator firmware require system rebooting, leading to system downtime and potential service disruptions, especially in applications that require continuous operation like large database systems.

Innovation Solution

A method for updating firmware on a coherent hardware accelerator during system runtime by storing the firmware update package in local memory, pausing operations, applying the update, and resuming them once complete, allowing for uninterrupted service.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the system reboots to apply firmware update on accelerator, then the firmware update is applied successfully, but system downtime occurs and service is disrupted

Engineering Contradiction:
Improvefirmware update reliabilityVSAvoidsystem downtime
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The firmware image is downloaded and stored in local memory on the accelerator before the update is applied. This preliminary preparation allows the update to be installed without requiring system reboot, as the new firmware is already available in local storage and can be activated by resetting only the accelerator component.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system separates the accelerator into an updatable component with its own local memory storage. This segmentation allows the accelerator to be updated independently from the rest of the system, enabling firmware updates without full system reboot and minimizing service disruption.

Inventive Principle:
Principle #1Segmentation

2Reliability

If the system reboots to activate firmware update, then the updated functionality is activated, but continuous operation is interrupted

Engineering Contradiction:
Improvefirmware update activationVSAvoidcontinuous operation duration
Core Design Contradiction:
ReliabilityVSDuration of action of stationary object

Solution Approach 1:

The firmware update package is downloaded and stored in local memory on the accelerator before the update is activated. This preliminary action eliminates the need for system reboot to load the firmware, allowing continuous operation to proceed uninterrupted while the accelerator is updated.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Local memory on the accelerator serves as an intermediary storage medium between the firmware update source and the accelerator firmware. This intermediary allows the update to be applied without requiring the operating system to reboot, maintaining continuous operation while enabling firmware activation.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Speed

If direct memory access is used by accelerator, then processing speed is improved, but memory management complexity increases

Engineering Contradiction:
Improveprocessing speedVSAvoidmemory management complexity
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

A memory management unit acts as an intermediary between the accelerator and system memory. This mediator handles the complexity of memory address translation and management, allowing the accelerator to use direct memory access for high-speed processing while the memory management unit manages the complexity of virtual to physical address translation and memory coherence.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS10671369B2Replacing an accelerator firmware image without operating system reboot
Publication Date: 2020.06.02 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US10671369B2 patent drawing
  • US10671369B2 patent drawing
  • US10671369B2 patent drawing

AI summary

The present disclosure relates to updating a firmware image on a coherent hardware accelerator concurrently with executing operations on the coherent hardware accelerator. According to one embodiment, while executing accelerator-enabled operations on the coherent hardware accelerator, a system stores a firmware update package in a local memory on the coherent hardware accelerator. Once the firmware update package is stored in local memory on the coherent hardware accelerator, the system restarting the coherent hardware accelerator by pausing the execution of at least a first operation initiated on the coherent hardware accelerator and applying the firmware update package to the firmware image on the coherent hardware accelerator. Once the firmware update package is applied to the coherent hardware accelerator, the system resumes the operation on the coherent hardware accelerator.