Diva: The "Holy Grail" of Range Filters with Navid Eslami (VLDB'25 Best Research Paper Award) скачать в хорошем качестве

Diva: The "Holy Grail" of Range Filters with Navid Eslami (VLDB'25 Best Research Paper Award)

In this episode of Disseminate: The Computer Science Research Podcast, host Dr. Jack Waudby interviews Navid Eslami (University of Toronto) about his VLDB Best Research Paper — DIVA: Dynamic Range Filter for Variable Length Keys and Queries. We explore how dynamic range filters are reshaping modern data systems, enabling faster queries, lower false-positive rates, and smarter indexing for real-world workloads. Discover how this breakthrough combines theory and practice to achieve what some call the holy grail of data filtering. What you’ll learn: What filters and range filters are — and why they matter How DIVA overcomes major challenges in modern databases The role of DIVA in WiredTiger (MongoDB’s storage engine) Dynamic data structures, query optimization, and practical takeaways for engineers How this research narrows the gap between database theory and applied systems Chapters: 00:00 – Intro: Welcome to Disseminate 00:40 – Meet Navid Eslami (University of Toronto & VLDB Best Paper Winner) 01:20 – What Are Filters in Databases? (Bloom Filters & LSM Trees Explained) 05:45 – From Filters to Range Filters — The Next Step 07:20 – Real-World Example: Range Queries in Databases 08:10 – Introducing DIVA: Dynamic Range Filter for Variable Length Keys and Queries 09:30 – The Problems with Existing Range Filters 12:40 – The Five Missing Ingredients of Current Solutions 15:20 – How DIVA Works: Sampling, Compression & the “Infix” Concept 21:00 – Handling Range Queries Efficiently 26:00 – Making DIVA Dynamic: Inserts, Deletes & Expansions 32:00 – Performance Evaluation & Benchmarks 35:10 – Integrating DIVA with WiredTiger (MongoDB’s Storage Engine) 38:40 – Open Source & Real-World Availability 40:30 – Surprises During Research — The Smooth Distribution Assumption 43:30 – Limitations and When NOT to Use DIVA 46:00 – Combining DIVA and Memento for Optimal Filtering 47:40 – The Future of Range Filters & Open Research Directions 49:30 – Practical Impact: What DIVA Means for Industry & Databases 51:30 – Final Thoughts for Practitioners and Researchers 53:40 – The Story Behind the Name “DIVA” 55:00 – Where to Find Navid Eslami & Outro Ideal for database engineers, systems researchers, and computer science students who want to understand the future of data systems. Resources & Links: Diva: Dynamic Range Filter for Variable Length Keys and Queries (VLDB'25) https://www.vldb.org/pvldb/vol18/p392... Diva’s GitHub: https://github.com/n3slami/Diva Navid's LinkedIn: https://www.linkedin.com/in/navid-esl...