At you 0:26 you describe write latency across continents for non active-active databases. In Torso if when you write to a local London replica (of Sydney), the write is still redirected to Sydney so torso still has the same latency issue, right? If so how is it different than a REST POST call from London to Sydney?
I love how you used deno in the video but the only problem is deno with drizzle is not the best experience. I made a simple cli to setup drizzle with deno but it'd be cool if drizzle also supported deno Edit: I am talking about drizzle kit specifically the drizzle lib is platform agnostic and is nice and works super well with deno
Embedded Replicas might be interesting also you’re still using their cloud to sync, otherwise hosting libsql yourself with a custom sync layer is think
Does the example code in this video lead to a database that is on the edge / not all in one place? I guess it does this automatically for read replicas, but not write replicas right? I was imagining something more like one database for every couple users located close to them, like in an online game for example (like Cloudflare durable objects). I think Turso mentions on their website "you could have one database per user", maybe explaining that could be another video idea
Thank you for the suggestion! Yes - you automatically get replicas, and you can add as many location as you want. Reads will always be routed to the closest location. Writes will always be routed to the primary, regardless of the user's location.
You are right, you could do Postgres via Supabase if you are into Backend as a Service solutions. The main selling point for Turso is its distributed nature and the fact that you can seamlessly add replicas at the edge.
Yesterday they increased the limit on number of databases to like 500 or so. Thanks for the video, I will try this
Haha I saw that!
In theory we care more about the total storage. I feel like 9GB is more than enough for a decent sized app.
@@awesome-coding yes 9gb is enough for proof of concept (non AI one's)
It would be interesting to see the cost and latency comparison with a more traditional setup.
I'm working on some tests for this exact scneario.
@@awesome-coding Great effort! Thanks, did you publish the results of the tests?
Time to try hono + cloudflare workers + torso.
LIVING ON THE EDGE
Nice stack! Will explore it!
Would love to explore how to sync a turso database with a client libsql db for extreme performance
A good idea for a future video! Thanks!
At you 0:26 you describe write latency across continents for non active-active databases.
In Torso if when you write to a local London replica (of Sydney), the write is still redirected to Sydney so torso still has the same latency issue, right? If so how is it different than a REST POST call from London to Sydney?
Think the latency benefit is mostly for reads, users usually “expect” some wait time when doing something, but browsing should be as fast as possible
I love how you used deno in the video but the only problem is deno with drizzle is not the best experience. I made a simple cli to setup drizzle with deno but it'd be cool if drizzle also supported deno
Edit: I am talking about drizzle kit specifically the drizzle lib is platform agnostic and is nice and works super well with deno
Hmm.. interesting. I'll check out drizzle kit.
When I hear TypeScript I think Deno these days.
@@awesome-coding lol me too. Bun if I have to work on a node project and deno for everything else
Okay, okay - can I run this on premises, my own servers or VPS?
I think you can install and deploy LibSQL on your own, but you'll miss the whole edge angle that Turso is offering.
Embedded Replicas might be interesting also you’re still using their cloud to sync, otherwise hosting libsql yourself with a custom sync layer is think
I think turso seems really interesting. I'm going to use it for my next personal hobby project to try it out
Glad to hear! Good luck!
You can push you schema written with drizzle-orm using drizzle-kit, avoids rewriting anything
You are right. Thanks for mentioning it!
Does the example code in this video lead to a database that is on the edge / not all in one place? I guess it does this automatically for read replicas, but not write replicas right?
I was imagining something more like one database for every couple users located close to them, like in an online game for example (like Cloudflare durable objects). I think Turso mentions on their website "you could have one database per user", maybe explaining that could be another video idea
Thank you for the suggestion!
Yes - you automatically get replicas, and you can add as many location as you want. Reads will always be routed to the closest location.
Writes will always be routed to the primary, regardless of the user's location.
Hey... Is your name Lucian opera? I've watched your PGSQL optimisation playlist...
Hey! It's not, but I'm also Romanian, so we all share the same awful accent :))
@@awesome-coding oh... my bad ;)
Thank you
Welcome!
at what scale does this become useful
I believe you can benefit from lower latencies at any scale.
at this point why not postgres?
You are right, you could do Postgres via Supabase if you are into Backend as a Service solutions.
The main selling point for Turso is its distributed nature and the fact that you can seamlessly add replicas at the edge.
cool