Join us virtually on Tuesday, June 10th for "Syncing Postgres into the Client" with James Arthur.
Postgres has binary and logical replication for syncing data between replicas. But what about syncing into the client? What are the challenges and requirements of extending replication all the way into the client? How can you make this simple, scalable, and secure without making the database a bottleneck or exploding resource use when scaling to lots of concurrent users?
This talk reports on our investigations building ElectricSQL, an open source Postgres sync engine. It covers topologies for client sync, pros and cons of different replication protocols, the need for partial replication, and bottlenecks we encountered for data throughput and scalability. It will demonstrate the results, showing 10 million concurrent clients syncing data in real-time from a single commodity Postgres with low and flat latency, CPU, and memory use.
The talk will then reflect on use-cases for this type of sync architecture and how it aligns with modern trends in software development. Particularly (a) how Postgres is being used with sync to build local-first applications with a high-quality, Figma, and Linear-grade UX and (b) how Postgres sync can be used to build local AI applications with low-latency data retrieval.
It will end by touching on some of the core limitations with snapshot and concurrent primitives that make it hard to build sync on Postgres, including current workarounds and ideas for improvements that could help to unlock further adoption of Postgres for these kind of sync-based use-cases.
Schedule:
12:00 PM Announcements, speaker introduction, presentation, Q&A.
1:15 PM Closing announcements, wrap up.
1:30 PM Event ends.
The link and detailed joining instructions, including the dial-in numbers for phone access, will be sent by email the day prior to the event, and posted on Meetup for those who have RSVPed "Yes".
General information about using GoToMeeting for this event can be found at:
https://bit.ly/GTMSFBayPUG
