Engineering Insights

Block Sparse Matrices for Smaller and Faster Language Models

Consensus algorithms at scale: Part 2 - Rules of consensus

If you’re still catching up, you can find links to each article in the series at the bottom of this article. The Rules of Consensus YouTube Scale When we were running Vitess at YouTube, there were tens of thousands of nodes serving very high QPS. The scale out was in all dimensions: some of the shards had over fifty replicas. The topology was complicated with these nodes being spread out across multiple data centers. To make this work, we had to strike a balance between latency, availability and

Node.js 14.10.0 (Current)

Learn How Prisma Supports Database Transactions | Prisma

Learn how Prisma supports database transactions, the tradeoffs of different transaction patterns, and how Prisma's transaction capabilities have evolved.

Generative language modeling for automated theorem proving

Learning to summarize with human feedback

We’ve applied reinforcement learning from human feedback to train language models that are better at summarization.

On joining PlanetScale and the vision of open source database infrastructure

This is a personal account on why I joined PlanetScale to work on Vitess and PlanetScaleDB, and what I perceive Vitess can become in the MySQL open source ecosystem. Background I am a software engineer, drawn into the database world. I’ve been working with MySQL for two decades now, and am authoring or have authored various open source tools in the MySQL ecosystem: orchestrator for high availability, gh-ost for online schema migrations, freno for throttling and others. In the past decade I’ve wo

September 2020

Read more