Hybrid Checkpointing for Computing System Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current checkpointing methods in computing systems face challenges in balancing fault tolerance with overhead, particularly in global checkpointing which incurs high storage and communication costs, and RAID technologies do not efficiently manage transient and permanent failures across multiple devices.
Innovation Solution
Combining local checkpointing with erasure-code protection, such as RAID, to store checkpoints across multiple devices, reducing space and communication overhead by using parity protection and aggregating log buffer updates, thereby providing robust fault tolerance with lower resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If global checkpointing is used to improve fault tolerance, then system reliability is improved, but storage overhead and communication costs increase significantly
Solution Approach 1:
The patent segments checkpointing into local and global components. Local checkpoints are stored on each device individually, while global checkpoints are created only when needed for permanent failures. This segmentation reduces the frequency and volume of global checkpoint operations, thereby reducing storage overhead while maintaining reliability.
Solution Approach 2:
Instead of performing full global checkpointing continuously, the patent applies partial checkpointing only when necessary (e.g., when permanent failures are detected). This partial action approach reduces storage overhead by avoiding redundant full system state captures while still providing fault tolerance when needed.
2Reliability
If global checkpointing is used to protect against permanent failures, then system reliability is improved, but communication overhead increases
Solution Approach 1:
The patent divides failure protection into two segments: local checkpoints for transient failures (handled independently by each device) and global checkpoints for permanent failures. This segmentation ensures that communication overhead is incurred only when global coordination is necessary, rather than continuously for all failure types.
Solution Approach 2:
The patent applies different checkpointing strategies to different failure types: local autonomous checkpointing for transient failures and coordinated global checkpointing for permanent failures. This local quality approach optimizes communication overhead by matching the checkpointing intensity to the specific failure scenario.
3Reliability
If RAID is used to provide redundancy, then fault tolerance is improved, but it does not efficiently manage transient and permanent failures across multiple devices
Solution Approach 1:
The patent introduces dynamic failure management that adapts to different failure types. Transient failures are handled through local rollback to previous checkpoints without involving other devices, while permanent failures trigger global checkpoint coordination. This dynamic response optimizes failure management efficiency compared to static RAID redundancy.
Solution Approach 2:
The patent changes the operational parameters of redundancy based on failure type. For transient failures, it uses local checkpoint parameters with minimal coordination. For permanent failures, it switches to global checkpoint parameters with full coordination. This parameter adaptation allows efficient handling of different failure scenarios.
Data Source
AI summary
Systems, methods, and computer-readable and executable instructions are provided for computing system reliability. A method for computing system reliability can include storing, on one of a plurality of devices, a checkpoint of a current state associated with the one of the plurality of devices. The method may further include storing the checkpoint in an erasure-code group across the plurality of devices.