Arun Sharma Open Source Activity
In his blog post, Mehul Shah asks: So, what does this (LLMS) all mean for infrastructure like databases? I argue it has three implications: First, the most prosaic implication is that we’ll need to build databases for LLMs. (Higher throughput) The second and more interesting implication is that we will likely build databases with LLMs. (Rewrite/Custom Develop database code itself) ...
This is some commentary on Performance Hints doc from Jeff Dean and Sanjay Ghemawat from someone who has written a lot less code and was a hot headed heretic in the 90s and whose career overlapped with Jeff and Sanjay a few times over the last 25-30 years. I don’t disagree with anything Jeff and Sanjay say in the doc. Most of their advice is sound and will result in faster code without sacrifi...
Navigating the Volcano: Efficient Query Processing in Graph Databases The Volcano iterator model, a cornerstone of relational database query execution since its introduction in the early 1990s, operates on a simple yet powerful principle: each operator in a query plan acts as an iterator that produces a stream of tuples one at a time, invoked via methods like open(), next(), and close(). This ...
Packaging Graphs: A Tale of Two Sparse Cousins Everyone knows tar files! Perfect for shuttling data between filesystems like ext4 and xfs, preserving structure, ownership, and timestamps for seamless transfers—ideal for backups or migrations. ISO files, meanwhile, are the mount-and-go champs, letting you access contents directly without unpacking, like a virtual disk for quick reads or softwar...
This post is inspired by a recent VLDB paper about database normal forms from Couchbase people. Relational Database 101 In the 1970s, a fellow named Edgar F Codd made a bunch of rules so you could swap out one data structure from the disk with another data structure and continue to use the same application. He also wanted to minimize data duplication and inconsistency using these rules. 1...