What If Your Hard Drive Could Predict Its Own Failures?

Table of Links

Abstract and 1. Introduction

2. Motivation and design goals

3. Related Work

4. Conformal prediction

   4.1. Mondrian conformal prediction (MCP)

   4.2. Evaluation metrics

5. Mondrian conformal prediction for Disk Scrubbing: our approach

   5.1. System and Storage statistics

   5.2. Which disk to scrub: Drive health predictor

   5.3. When to scrub: Workload predictor

6. Experimental setting and 6.1. Open-source Baidu dataset

   6.2. Experimental results

7. Discussion

   7.1. Optimal scheduling aspect

   7.2. Performance metrics and 7.3. Power saving from selective scrubbing

8. Conclusion and References

7. Discussion

The proposed method for disk identification for scrubbing offers a dual benefit. Firstly, it can be utilized to assess the reliability of the storage system. Secondly, it employs a disk ranking mechanism to assign relative health scores to individual disks. The choice of classification algorithm depends on factors such as dataset size and available compute resources. However, the decision can be guided by the expertise of the system administrator.

In addition, we discuss how the use of the Mondrian conformal predictor can aid in identifying latent failures of disks, which could be a potential area for future research. Furthermore, we identify three key aspects for designing optimal scheduling and cover performance metrics, including effective coverage and size of the average prediction set.

Lastly, we provide a hypothetical evaluation of energy and power savings resulting from selective scrubbing. This showcases the potential benefits of the proposed method in terms of reduced power and energy consumption, highlighting its effectiveness in optimizing disk scrubbing operations.

7.1. Optimal scheduling aspect

With respect to disk scrubbing frequency scheduling, we can design three aspects of scheduling: time window, frequency, and space allocation. Each of them is described below:

• Time window focuses on scheduling the time window for scrubbing based on the workload pattern. Scrubbing is done when the system is predicted to be idle.

• Frequency involves scheduling the frequency of scrubbing based on the health status of the drive. For drives with the best health, scrubbing is done less frequently. For drives with medium health, scrubbing is done more frequently.

• Space deals with scheduling space allocation based on the spatial and temporal locality of sector errors. Instant scrubbing is performed on problematic chunks to ensure efficient disk scrubbing.