Bibliographic record
Abstract
A workflow management system is responsible for scheduling distributed computing jobs on a cluster, tracking their progress, and recovering from failures. Each job consists of a flow of tasks that form a dependency graph. Workflow management systems are becoming increasingly important as modern applications continue to grow in scale and distribution. This thesis begins by analyzing existing workflow management systems such as Celery and Ray, identifying several key limitations. Celery suffers from scalability issues due to high task scheduling latency and lacks support for task locality. Ray, on the other hand, does not support static task dependency graphs, resulting in significant startup overhead forchained tasks. Furthermore, both systems provide only limited support for failure recovery. To address these challenges, this thesis introduces the design and implementation of Spider, a workflow management system optimized for efficiently executing latency-sensitive tasks in distributed environments. Spider features a lightweight scheduler to reduce task scheduling overhead and improve scalability. It introduces a Data abstraction to track data locality, enabling intelligent task placement and minimizing unnecessary data transfers. The Data abstraction also plays a central role in failure recovery by allowing jobs to resume execution without redundant recomputation. In addition, it supports background garbagecollection – a capability lacking in existing systems – enabling efficient cleanup of temporary data without disrupting task execution. Through both microbenchmarks and real-world benchmarks, we demonstrate that Spider outperforms both Celery and Ray on large-scale, latency-sensitive workloads.
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 | 0.004 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.005 | 0.002 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.015 | 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".