Battery Learn Cycle Management via ML Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Battery backup modules in data storage systems require periodic learn cycles, which can disrupt data integrity due to power outages and increase latency during write-back caching, while write-through caching reduces data loss risk but incurs higher latency costs.
Innovation Solution
A computer-implemented method using a machine learning model to determine optimal time slots for battery learn cycles based on disk usage telemetry, minimizing the impact on disk I/O operations by predicting low I/O periods, thus reducing the latency associated with battery learn cycles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If write-back cache policy is implemented to improve write performance, then write speed is improved, but data loss risk increases due to potential power interruption before cache data is written to physical disk
Solution Approach 1:
The battery is pre-charged to full capacity before power interruption occurs, creating a ready power reserve that can immediately sustain the cache memory. This preliminary action ensures that when power failure happens, the cache remains powered long enough to flush data to disk, thus maintaining both high write performance and data safety without requiring synchronous writes.
2Reliability
If battery learn cycle is performed periodically to maintain battery capacity, then battery reliability is improved, but system latency increases during the learn cycle process
Solution Approach 1:
The patent implements dynamic scheduling of battery learn cycles by monitoring real-time disk I/O workload patterns. The system adaptively selects time slots with lowest I/O activity for learn cycle execution, and can dynamically adjust learn cycle frequency based on battery health status and system usage patterns. This dynamic approach minimizes latency impact while maintaining battery reliability.
Solution Approach 2:
The battery learn cycle is implemented as a periodic maintenance operation rather than a continuous process. The system schedules learn cycles at predetermined intervals based on battery usage, and can skip or delay cycles when system workload is high. This periodic execution with conditional scheduling reduces overall system latency while ensuring battery capacity is maintained at acceptable levels.
3Reliability
If battery learn cycle is performed to restore holdover time, then battery capacity is restored, but disk I/O operations are disrupted during the learn cycle
Solution Approach 1:
The patent applies different quality levels of service to different operations during battery learn cycle. Critical I/O operations receive higher priority and are serviced first, while non-critical operations are delayed or throttled. The system creates localized quality differentiation in I/O handling, ensuring that learn cycle execution does not uniformly degrade all disk operations but rather allows time-sensitive operations to proceed with minimal disruption.
Solution Approach 2:
The controller acts as an intermediary between the battery learn cycle process and disk I/O operations. It buffers and manages I/O requests during learn cycle execution, queuing non-urgent operations and prioritizing critical ones. This intermediary buffering mechanism decouples the learn cycle from direct I/O interference, maintaining overall system productivity while allowing battery maintenance to proceed.
Data Source
AI summary
In one aspect, an example methodology implementing the disclosed techniques includes, by a disk controller of a plurality of disks of a data storage system, collecting usage telemetry from the plurality of disks and initiating determination of one or more candidate time slots for performing a battery learn cycle using a machine learning model. The machine learning model is trained using the usage telemetry from the plurality of disks. The method also includes, by the disk controller, presenting one or more of the one or more candidate time slots for performing a battery learn cycle.