Front Matter

Sihem Amer-Yahia, Jian PeiPVLDB 2018

Proceedings of the VLDB Endowment

This publication represents the front matter for Volume 11, Issue 5 of the Proceedings of the VLDB Endowment (PVLDB). It curates a selection of high-impact research papers presented at the 44th International Conference on Very Large Data Bases (VLDB 2018), held in Rio de Janeiro, Brazil.

The PVLDB Publication Model

The "Letter from the Program Chairs" highlights the unique "journal-conference hybrid" model of PVLDB. Unlike traditional conferences, PVLDB operates with twelve monthly submission deadlines and a rigorous journal-style review process, including a three-month revision cycle. This ensures that the community receives "journal-grade" research with the rapid dissemination characteristic of a major annual conference.

Addressing Modern Data Management Challenges

The papers in this issue tackle several critical shifts in the field of very large databases:

1. Privacy and Utility

A major focus is the tension between data utility and user privacy. Research in this issue explores how to enable SQL analytics while maintaining strict guarantees of differential privacy, a necessity in the era of GDPR and heightened data sensitivity.

2. Hardware-Conscious Indexing

As hardware evolves, so must database internals. The BzTree is introduced as a solution for non-volatile memory (NVM), utilizing latch-free techniques to maximize concurrency on modern many-core systems without the bottlenecks of traditional locking mechanisms.

3. Machine Learning Infrastructure

The convergence of databases and machine learning (ML) is a recurring theme. The issue features systems like Ease.ml, which manages resource sharing for multi-tenant ML workloads, and FlexPS, which optimizes parameter server architectures. These contributions treat ML models as first-class citizens within data resource management frameworks.

4. Industrial Scale and the Cloud

Bridging the gap between academic theory and industrial practice, the issue includes an industry-track look at Apple's CloudKit. This contribution details the structured storage architecture powering massive mobile ecosystems for hundreds of millions of devices.

VLDB 2018 Conference Banner

Technical Methodologies and Insights

The issue highlights diverse technical approaches to data problems:

  • Latch-free Indexing: Maximizing throughput on modern hardware.
  • Programming-by-Example (PBE): Automating the migration of hierarchical data into relational tables to reduce manual re-architecting effort.
  • Adaptive Indexing: The TOAIN system provides throughput-optimized indexing for dynamic kNN queries on road networks.
  • Multi-Truth Discovery: Algorithms designed to resolve conflicting information from multiple sources using domain-aware context.

Conclusion

Spanning pages 526 to 662, this issue of PVLDB exemplifies the ongoing evolution of data management. By integrating privacy, optimizing for heterogeneous hardware, and building the infrastructure for large-scale machine learning, the research presented here sets the stage for future innovations in the "Very Large Data Base" domain.

Summary updated

BibTeX
@article{ameryahia2018front,
  author = {Sihem Amer-Yahia and Jian Pei},
  title = {{Front Matter}},
  journal = {Proc. VLDB Endow.},
  year = {2018},
}
Figure