Planetscale would support scaling and replication from the start, D1 is still building it
Planetscale would support scaling and replication from the start, D1 is still building it
MNo but cool idea - best is probably logging the prepared query string
wrangler d1 insights could tell you summary stats but it only shows the query string too
Aregardless of that technically speaking, SQLite would be worse?
Mthere's no better/worse here, It Depends
Aon?
Mquery execution time inside D1, yeah
Mexpected reads/writes per second and amount of data stored to start with - sqlite can be really good for scaling simple read-heavy use cases really far as there's less to worry about
Mwhereas the tech underneath planetscale powers youtube, but you pay a price for that scalability
Athanks for the run down
Ais there a simple reason why CF chose SQLite then since mySQL seems superior
Msimple reason? easiest, most battle-tested path to enable an sql interface for storing data for Workers
Agot it
Aill just stick with keeping everything under cloudflare for now then and down the line when i do need "more" if CF isnt enough ill look into Planetscale
Aim reading the docs to query D1 locally with bindings, but im using next-pages and so when i develop locally i just use npm run dev for fast refresh, how am i supposed to initialize the prisma client in that case?
Athis is what i was using before for planetscale but obviously no database_url for d1
TIt's hard to maintain a large mysql cluster for the edge. Even if they do, the price will be as high as everest
TIn my case, D1, KV and R2 is perfectly fit. Most of the time my request are pagination, filtering with some predefined selected columns and get by ID. In mysql I will need a DB proxy to reduce the connections created and cache requests. With D1, it's not really faster but it's much cheaper without the worrying of connection managing and query caching.
Tyeah there are no dynamic binding yet
Tbut having a 50k warming database at the same time is mindblown
MLooks like we still need to manually configure database IDs in a global wrangler file as there’s no programmatic way to manage a fleet of D1 instancesYeah, dynamic bindings hasn't shipped yet, there's an unofficial package to automate the HTTP API steps you'd need to do atm though https://github.com/nora-soderlund/cloudflare-dynamic-bindings
Mwhat does D1 per DO mean
HSQLite Storage Bindings for DOs?
Mnot sure, unrelated work
MYou aren't going to get D1's in DO though, you'd get a raw sqlite instance
HWould the SQLite instance support replication?
HThat would be metal...
Monly if DO itself did
HAren't they two different storage systems?
HOr...
Mthink of D1 as the opinionated way to build a database service on top of raw sqlite+DO
HOr something like Code Procedures?
HOr whatever that "Colocate-your-code-next-to-D1" thing was called
M?
Moh you're asking if DO will support things we announced for D1
Mno idea
HIn one of the earlier blog posts for D1, there was something talking about being able to colocate some code next to a D1 instance. Not sure if that is actually a DO, or Service Binding, or something else
JTrying to determine if D1 is right for a current project...
For use cases where the size of a single table could grow quite large (think hundreds of thousands of rows), is D1 simply not a good option due to the 10GB limit?
For use cases where the size of a single table could grow quite large (think hundreds of thousands of rows), is D1 simply not a good option due to the 10GB limit?
Mwe have dbs with millions of rows, no dramas
Jthis is reassuring to hear... so basically I shouldn't ever have to worry about it (assuming i go with a per-user approach).
Jfor the dbs with millions of rows, what's the approximate size?
Ji've seen postgres dbs at 300 to 400 GB... that's my only context for this... maybe sqlite is inherently smaller or something?
M2ish GB
MWorkers also supports big classic postgres DBs - using Hyperdrive
JThis is all very helpful. Thank you.
Yeah, no base64 images or anything, mostly just simple String and such.
Yeah, no base64 images or anything, mostly just simple String and such.
Mwe picked sqlite particularly because we could load millions of dbs across the world really quickly - that pattern breaks down for extremely large, monolith-style DBs
Jmakes sense... trying to work with 300-400 GB DBs is also just flat-out painful...
J(from my experience at least)
Athis makes sense, thanks
Ais there a guide on setting up d1 with prisma edge in local development?
