Tag
An introductory article describing Bε-trees as write-optimized key-value data structures, comparing their asymptotic and real-world performance against B-trees and LSM-trees for on-disk storage in databases and file systems.
A Pivot blog deep-dive explaining why 100% CPU utilization in databases often means the CPU is stalled on memory accesses, demonstrating how naive HashMap-based GROUP BY aggregations fail to exploit CPU caches and how much faster in-cache data processing can be.
ParadeDB details how segment partitioning (based on the Spooky LSM compaction paper) and other optimizations dramatically sped up COUNT queries and filtered, sorted searches—up to 66x faster than before—on Postgres-based search workloads.
Jim Calabro introduces One Hash Bloom Filters, a variant that reduces hashing overhead in bloom filters, building on double-hashing techniques and Bluesky's production experience where bloom filters save millions of database queries per second.
PostgreSQL 19 的发布被推迟至 11 月 19 日，原因是 LLM 辅助代码审查发现了大量缺陷，导致 8 到 10 个重要功能（包括 SQL/PGQ 图查询支持）被回滚到 PostgreSQL 20。
Sergio Pineiro 惊叹某开源搜索工具能在 30 毫秒内检索 3200 万条记录，性能极其惊人，并且设置简单、支持模糊搜索、向量搜索和语义搜索。
Materialize describes how it reworked its `persist` metadata store to build a linearizable log using PostgreSQL's READ COMMITTED isolation instead of SERIALIZABLE, tripling write throughput after a metastable latency incident, and validated correctness by modeling PostgreSQL's behavior in Lean 4 to prove linearizability.
Anton Zhiyanov (@ohmypy) 报道 SQLite 核心团队正在开发 Vec1——SQLite 原生向量搜索（ANN）功能，支持欧氏（L2）与余弦距离，并使用 SIMD 加速向量运算；该功能已基本完成，仅剩测试改进等工作。
论文提出 ARCS（Ambiguity Resolution Corpus for SQL），首个包含真实数据库上自然歧义的 text-to-SQL 基准，并提出“结构化消歧”范式取代自由对话澄清。实验表明歧义下 text-to-SQL 仍具挑战：GPT-6 SOTA 仅 51% 端到端执行准确率，开源模型均不超过 27%。
CedarDB ported the original 1993 Doom's game logic and BSP-based renderer entirely into SQL queries, running the game loop at 35 FPS with a full playable deathmatch demo inside the database.
Lukas Vogel's SQLDoom project renders full-color 640×480 Doom frames at 35 FPS using an SQL database, with game logic implemented in about 1,300 lines of SQL across 89 common table expressions, building on his earlier DoomQL project.
A partially unsealed court filing alleges DHS/ICE agents used Palantir's Investigative Case Management database to build dossiers on ICE protest observers, including photos, license plates, and facial recognition searches, while allegedly intimidating First Amendment observers.
Peter Mattis (CTO/cofounder of Cockroach Labs) shares a thread explaining B-tree node layout and spatial locality, inspired by recent B-tree discussions on the Pragmatic Engineer podcast.
A technical deep-dive explaining how traditional lock coupling in concurrent binary trees suffers from physical lock contention at the root node, limiting scalability as CPU core counts grow, with benchmark results on a 16-core system.
A technical deep-dive blog series exploring join ordering in query optimization, focusing on the German school of dynamic programming approaches (DPsize, DPsub, DPccp, DPhyp) and related work on adaptive optimization and join search space enumeration.
GreptimeDB 1.2 introduces JSON2 columns in all ingester SDKs, enabling storage of hot JSON paths as separate columns for more efficient querying and reduced resource usage compared to the previous JSONB type.
This post discusses the limitations of traditional databases for modern applications and promotes converged databases as a unified solution to handle diverse data types like relational, JSON, and vector data in a single system.
A comprehensive guide to core system design building blocks, including client-server architecture, scaling, and databases, to help with system design problems and interview preparation.
The tweet predicts that the future of software will be shaped by technologies like Postgres, Turbopuffer, Jev, Clickhouse, and open-source LLMs.
This academic paper revisits the dataflow model, likely providing new insights or updates on its theoretical foundations and applications in data management systems.