Data Mover Offloads Application Checkpointing from Processing System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional application checkpointing systems in information handling systems require multiple processing cycles for each snapshot operation, especially exacerbated by applications and server devices that create repeated checkpoint versions, leading to inefficiencies.
Innovation Solution
An Information Handling System (IHS) with a data mover processing system and memory system that performs application checkpoint and recovery operations by capturing and storing application information in a memory fabric, offloading these operations from the processing system to a data mover device, allowing for efficient storage and retrieval of application snapshots.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional application checkpointing systems utilize the processing system to copy application snapshots from local memory to persistent storage, then application fault tolerance is achieved, but processing system cycles are consumed and processing overhead increases
Solution Approach 1:
The patent extracts the checkpointing copy operation from the processing system and assigns it to a dedicated data mover device. The data mover independently copies application snapshots from local memory to persistent storage without consuming processing system cycles, thereby maintaining fault tolerance while preserving processing efficiency.
Solution Approach 2:
The patent introduces a data mover as an intermediary component between the processing system and storage systems. This intermediary handles all snapshot copy operations, acting as a buffer that protects the processing system from I/O overhead while ensuring reliable data transfer for checkpointing.
2Reliability
If applications create repeated checkpoint versions, then application recovery capability is enhanced, but the number of copy operations increases and processing overhead is exacerbated
Solution Approach 1:
By extracting the copy operation from the processing system to a dedicated data mover, multiple checkpoint versions can be created without each operation consuming processing system cycles. The data mover handles repeated copy operations independently, enabling frequent checkpointing without time loss to processing overhead.
Solution Approach 2:
The data mover is pre-configured and ready to perform copy operations as soon as snapshots are generated. This preliminary preparation allows checkpoint operations to execute immediately without waiting for processing system availability, reducing the time penalty for creating multiple checkpoint versions.
3Measurement precision
If the processing system freezes or stops application operation to capture snapshots, then accurate application state information is obtained, but application execution time is interrupted
Solution Approach 1:
The data mover acts as an intermediary that can capture and copy application state information from memory without requiring the processing system to freeze or stop the application. This intermediary approach maintains application execution continuity while still obtaining accurate snapshots through direct memory access.
Solution Approach 2:
The patent replaces the mechanical approach of freezing application execution to capture state information with a non-intrusive method where the data mover directly accesses memory during normal operation. This substitution eliminates the need to interrupt application flow while maintaining snapshot accuracy.
Data Source
AI summary
An application checkpoint and recovery system includes an application provided using application information from a first location in a memory fabric. A data mover device receives an application checkpoint instruction and performs application checkpoint operations including capturing a first version of the application information from the first location in the memory fabric, and providing the first version of the application information in a second location in the memory fabric. The data mover device may receive an application recovery instruction subsequent to the application modifying the first version of the application information to provide a second version of the application information in the first location of the memory fabric, and may transfer the first version of the application information from the second location in the memory fabric to the first location in the memory fabric so that it is available for use in providing the application.