Note: This is a draft I never finished. Some important topics may be left out.
Today on stream, Theo talks with Sam from Planetscale about scalability and performance in serverless database solutions like Planetscale.
Development
DynamoDB and Scalability
Sam and Theo talk about how DynamoDB locks its users in by making it difficult to move off the platform.
- AWS Acronym about keeping customers:
Once they have these five things they never leave. DynamoDB is one of them- Sam
Performance
- New features, compounding performance as a first priority
Reliability
- Sam: “Some of the greatest outages in my career have come from Redis”
Scalability of Pscale
-
Theo: Even vercel won’t scale for ping forever
-
Sam: “It’s my goal that you never have to hire that database person”
-
Theo: We just added a feature where we’re writing to our DB every 30 seconds per user, and we don’t have to worry about that with Planetscale.
DX
- Theo: planetscale is a lot like astro in that it’s bridging a gap between DX and perfoemance
- Sam: “No dev likes the experience of waking up at 2 in the morning”
Why MySQL instead of Postgres
- Sam: “Postgres is a fine database, but you can’t operate it as easily as MySQL”
- Sam: Way more companies moving Postgres -> MySQL than the other way around
Planetscale’s thoughts about limits
- Sam: Connection limits plague a lot of people. We aim to mitigate that so you don’t have to worry about it.
- Theo: You can set up PGBouncer or just use Planetscale, it’s that easy.
Database-js and HTTP database access
- Sam: We’ve got a lot of requests to make it easier to not use a driver
- Sam: Database-js makes it trivially easy to get your database connected, this is easier to use in, i.e. Cloudflare workers. We want to make it easy to leverage Planetscale
Misc
- Serialization is horrendous for performance in distributed systems
- Hobby projects need to scale too. (Pscale is hobby-friendly)
- Theo: Edge data is going to be one of the biggest new paradigm shifts in the industry.
Business
Graduation Curves
- Customers can become pitfalls or be stepping stones
- Vercel is trying to keep people in house
Industry Adoption
- When a tool is really good, the industry will change around it to adopt it. See figma.
- Sam: “Github didn’t plan on being the world’s platform for code”
Misc:
- Theo: “A team of 5 passionate people will beat a team of 100 people who lucked into their success” Sam: “I want to build a company where being passionate drives you forward.”
- The Lunch Table conversation, community is important.
- People are starting to trust the cloud more
Kubernetes Rant
Takeaways:
- Kubernetes is hard to maintain, so need an engineer like primeagen
- Very low use case anyway
- Better to use existing service for most apps that are small enough, as the cost would be lower than hiring an engineer