Technical explorations of database systems research, focusing on architecture, optimization, and system design through the lens of doctoral research.
Semantic query processing is emerging as a new layer atop relational engines, elevating LLM-backed semantic operators to first-class SQL primitives for multimodal data. We present SemBench, the first benchmark to rigorously evaluate these systems end-to-end, and outline our roadmap towards our own system, Spectra, to make semantic operators affordable at scale.
Performance optimizations of `ORDER BY ... LIMIT k`, also known as Top-K queries in SQL, in the Snowflake Data Warehouse. Joint work with Juliane Waack. Part 2.
Performance optimizations of `ORDER BY ... LIMIT k`, also known as Top-K queries in SQL, in the Snowflake Data Warehouse. Joint work with Juliane Waack.
I had the pleasure the work on TiKV, a distributed key-value store in Rust that serves as a storage engine for TiDB, as an LFX Program Mentee in 2021. The goal was to implement a "pluggable" coprocessor, similar to HBase's coprocessor, where users can leverage the computational power of storage nodes and directory execute arbitrary computation on them.
Automatically generate performance flamegraphs for Rust benchmarks with Criterion using pprof and in-process profiling hooks.
Sometimes, especially with the factory pattern, you want your derived classes to register automatically to a factory class. However, I found this harder than expected! Especially if you want it to be header only. Here is a quick walk-through of how this can be done.
RapidJSON is a nice C++ library for parsing JSON files. If you use CMake, we can automatically download it as a dependency for your project.
While .Net projects usually require a runtime environment, the Mono project allows us to compile a program to native code with all dependencies included. In this blog post, we will explore how to do this.