bohanoai

3 ms·

1. 1.▲
   by bohanoai 1y ago

   > The key insight, and main message, of the talk was that if you are not too write heavy you can scale Postgres to very high read throughput with read replicas and only a single master! That is exactly the message that needs to be spelle
2. 2.▲
   by bohanoai 1y ago

   Speaker here — Bohan from OpenAI. Our application has hundreds of endpoints, which makes sharding non-trivial. We've already offloaded shardable workloads—particularly write-heavy ones—from PostgreSQL. What remains is primarily read-on
3. 3.▲
   by bohanoai 1y ago

   Speaker here — Bohan from OpenAI. Just to clarify, we're using Azure Database for PostgreSQL, not a self-hosted setup. I mentioned "Azure Postgres" multiple times during the talk, but I should have been more specific that we&