ClickHouse is an open-source, column-oriented database for real-time analytics. Written in C++ and ships as a single binary. Queries run on a vectorized execution engine, and the MergeTree family of table engines is the main storage format.
ClickHouse started in 2009 at Yandex as a filter and aggregation operator for web-scale log data, the click records of the Yandex.Metrica web analytics service. It was open-sourced in 2016.
How Apache Kafka works internally, making sense of its many configurations across the event stream, brokers, and best practices for running a...
Understanding BFS and DFS graph traversal, the difference between visiting and exploring a node, with implementations and examples.
A deep dive into the R-tree, the data-driven spatial index built on bounding rectangles, and how it powers range and nearest-neighbor queries....
A detailed breakdown of Apache Doris, the MPP real-time data warehouse that speaks the MySQL protocol, covering its storage engine, meta service,...
Tracing a single Flink job end to end through the Client, JobManager, and TaskManagers to understand Flink's architecture, choosing depth over breadth....
Implementing breadth-first search with a stack instead of the usual queue, and what changes when you flip the underlying data structure.
A diagram-first introduction to graph theory from a computer science angle, covering what graphs are, the different kinds, and their real-world uses....
Tessellation for spatial indexing divides space into non-overlapping shapes for efficient data management. The Uber H3 grid system uses hexagonal cells, offering...
Grid systems in spatial indexing, including Geohash and Google S2, encode geographic locations for efficient spatial data management. Geohash uses a Z-order...
Spatial indexes optimize multi-dimensional data querying using space-filling curves like Z-order and Hilbert curves. Z-order curves employ a recursive Z pattern, and...
How B-trees and B+ trees organize data on disk for fast lookups, starting from disk structure and why databases lean on them...
Modeling and querying graph data in PostgreSQL when a dedicated graph database would be overkill, starting from graph fundamentals.
What a cartogram is and how it reshapes geography by statistical data such as population, laying the conceptual groundwork for building one....
A reflection on why challenging existing solutions and reinventing the wheel drives real progress, despite the popular advice against it.
How distributed training accelerates deep learning across many processors, using data parallelism and model parallelism to cut training time.
My accepted Google Summer of Code 2022 proposal for Our World in Data, building a population-scaled, shape preserving world cartogram.
Introducing hybrid quad-kd and r-kd tree structures that combine the quadtree, KD-tree, and R-tree to speed up search over mixed point and...
Why geometric algorithms like the convex hull break under floating point arithmetic, and the round off errors that produce non-convex or never-ending...
How external merge sort handles datasets too large for memory, using a split phase and a priority queue driven K-way merge, walked...
Replacing sprawling switch statements over enums with the Visitor pattern in Java, for code that stays clean and easy to extend.
Building MYRO, a myoelectric prosthetic arm that reads muscle activity through EMG sensors and pattern recognition to drive a 3D printed hand....
A four-part field guide to multi-dimensional indexing, read best in order.