←back to thread

Dolt is Git for data

(www.dolthub.com)
358 points timsehn | 2 comments | | HN request time: 0s | source
1. hypewatch ◴[] No.22735014[source]
Does Dolt have any benchmarks against other databases at scale? I would think that a git SQL database would not be very snappy at scale
replies(1): >>22735045 #
2. timsehn ◴[] No.22735045[source]
We're working on building performance benchmarks right now. We started with correctness. You can read about our correctness journey here: https://www.dolthub.com/blog/2019-12-17-one-nine-of-sql-corr...

We think over time (like years) we can achieve read performance parity with MySQL or PosgreSQL. Architecturally, we will always be slower on write than other SQL databases, given the versioned storage engine.

Right now, Dolt is built to be used offline for data sharing. And in that use case, the data and all of its history needs to fit on a single logical storage system. The biggest Dolt repository we have right now is 300Gb. It tickles some performance bottlenecks.

In the long run, if we get traction we imagine building "big dolt" which is a distributed version of Dolt, where the network cuts happen at logical points in the Merkle DAG. Thus, you could run an arbitrarily large storage and compute cluster to power it.