Bibliographic record
Abstract
Online transaction processing (OLTP) databases are a critical component of modern computing infrastructure services. As a result, they must be highly available, and they must process requests efficiently for a wide range of workloads. Databases provide high availability by using replication so that when a database replica fails, a backup can take over. They use workload management mechanisms for efficiently supporting different types of workloads with varying levels of skew and contention. This thesis revisits the challenges of database replication and workload management for in-memory databases. Unlike traditional disk-based databases, in-memory databases are designed for workloads whose entire dataset fits in DRAM memory. These databases are highly scalable, raising challenges for replication and workload management. For example, traditional database replication schemes suffer from the network, instead of storage, bottlenecks because in-memory databases have much higher throughput, and traditional workload management solutions significantly limit the performance of in-memory databases. In this thesis, we propose using deterministic concurrency control as the basis for replication and workload management. Deterministic concurrency control allows transactions to execute concurrently while guaranteeing equivalence to a predetermined serial ordering of transactions. For data replication, we propose a replay-based scheme that executes transactions concurrently and scalably on the backup database in the serial order predetermined by the primary database. Our solution reduces network bandwidth requirements to 10-15% of traditional database replication schemes. For workload management, we propose two optimizations to deterministic concurrency control that help parallelize internal database operations. These optimizations enable handling contention and skewed workloads efficiently and provide 30% to 6x performance improvements.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 source (direct Gemma or distilled Codex), 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".