Bibliographic record
Abstract
The reliability of storage systems is critical for our modern world. As the world's demand for data expands, an ever increasing number of systems and devices are required to store and process these data. A single data center often contains hundreds of thousands and even millions of nodes. The numerous components dictate that failures within these clusters are no longer the exception but the norm, and simplistic assumptions about error characteristics and mitigation are insufficient at these scales. The cost of operation for individual systems as well as massive data centers are contingent on low rate of error incidences and successful, timely mitigation to minimize the impact to regular operations. To that end, this thesis studies the reliability of different components across the storage stack and develops mitigation strategies based on insight gained from large-scale, real-world data. Specifically, we study main memory and disks, two of the most frequently replaced components in data centers. We analyse memory error characteristics from several large-scale systems and recommend simple page retirement policies to mitigate errors. We further develop a novel technique to protect previously unprotected pages that belong to the kernel, by repurposing virtualization hardware already available on modern processors. We also evaluate several new error mitigation mechanisms available on modern hardware. At the same time, we examine the effect that temperature has on system performance and power draw. Lastly, we study the reliability of file systems on solid state drives by injecting errors based on realistic SSD fault modes. We discover bugs that are present in popular file systems and make recommendations on future file system design. Overall, we advance the understanding of error characteristics and mitigation in several aspects and across different layers of the storage stack. We study errors by observing systems in deployment, and use the insights gained to develop effective mitigation that target real-world scenarios.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".