3 ms·

> If you can figure out how to use D1 as a frontend dev, you can use a hosted Postgres solution and get much more power and performance. After a cursory glance

> If you can figure out how to use D1 as a frontend dev, you can use a hosted Postgres solution and get much more power and performance. After a cursory glance at Cloudflare D1's description and service limits, I was left with the impression that the usecases are not that of a generic RDBMS. I mean, the limit for each database is 10GB, and apparently a D1 database resides in a single location. Of course this means >200ms to even hit the server when requests don't come from the same data center. Perhaps this is one of those things where a underperforming implementation is released to minimize time-to-market, and then they work on performance. However, for a distributed service this suggests to me that this solution only works as an ad-hoc storage of complex relational data for high-read, low-write scenarios, where you have Cloudflare Workers query D1 but either aggressively cache the response or store queries in Cloudflare KV to avoid the roundtrip cost.

fastball 2y ago
It doesn't even work well for high-read/low-write scenarios, that was my issue with it. If I'm caching everything in Cloudflare KV anyway, why would I want to use a terribly performing SQL store for cold operations? Much better to just use a hosted postgres and cache things in CF KV if that isn't fast enough.
arkh 2y ago
> Of course this means >200ms to even hit the server when requests don't come from the same data center. What world are you living in? When using a European server from Europe I expect < 10ms ping. Less than 100ms for East coast server, less than 150ms for west coast ones. Only cases I'd expect more than 200ms (and not a lot more) is when hitting Australian servers.
- motorest 2y ago
  > When using a European server from Europe I expect < 10ms ping. I'm living in reality, where a globally available application needs to serve people all over the globe from Cloudflare's >300 data centers. This means a worker running in the same data center hosting your D1 databases might get ~10ms requests, but workers running on all other >300 data centers have to pay for the round trip to the primary data center whenever there is a cache miss. The question you need to ask yourself is how come you failed to notice that. Do you have a good grasp on what you are doing?
  - arkh 2y ago
    > This means a worker running in the same data center hosting your D1 databases might get ~10ms requests This is already waaaaaay too much. Same datacenter I expect nano seconds round trips. Datacenter to a datacenter on the other side of the world? Maximum 300ms. It was what you expected of game servers two decades ago. I don't see what could eat more than a couple dozen ms on the server with current hardware to make a 10ms round-trip become "> 200ms" and consider this a service you'd pay for.
    - manter 2y ago
      I think D1 actually requires multiple round trips, so the higher figures start to make some kind of sense. I read a blog post about this but can't find it now. I did find this [release note](https://developers.cloudflare.com/d1/platform/release-notes/#2025-01-07 https://developers.cloudflare.com/d1/platform/release-notes/...) though. They describe removing _at least_ two round trips back to the DC hosting the D1 instance.

fastball 2y ago

It doesn't even work well for high-read/low-write scenarios, that was my issue with it. If I'm caching everything in Cloudflare KV anyway, why would I want to use a terribly performing SQL store for cold operations? Much better to just use a hosted postgres and cache things in CF KV if that isn't fast enough.

arkh 2y ago

> Of course this means >200ms to even hit the server when requests don't come from the same data center. What world are you living in? When using a European server from Europe I expect < 10ms ping. Less than 100ms for East coast server, less than 150ms for west coast ones. Only cases I'd expect more than 200ms (and not a lot more) is when hitting Australian servers.

motorest 2y ago
> When using a European server from Europe I expect < 10ms ping. I'm living in reality, where a globally available application needs to serve people all over the globe from Cloudflare's >300 data centers. This means a worker running in the same data center hosting your D1 databases might get ~10ms requests, but workers running on all other >300 data centers have to pay for the round trip to the primary data center whenever there is a cache miss. The question you need to ask yourself is how come you failed to notice that. Do you have a good grasp on what you are doing?
- arkh 2y ago
  > This means a worker running in the same data center hosting your D1 databases might get ~10ms requests This is already waaaaaay too much. Same datacenter I expect nano seconds round trips. Datacenter to a datacenter on the other side of the world? Maximum 300ms. It was what you expected of game servers two decades ago. I don't see what could eat more than a couple dozen ms on the server with current hardware to make a 10ms round-trip become "> 200ms" and consider this a service you'd pay for.
  - manter 2y ago
    I think D1 actually requires multiple round trips, so the higher figures start to make some kind of sense. I read a blog post about this but can't find it now. I did find this [release note](https://developers.cloudflare.com/d1/platform/release-notes/#2025-01-07 https://developers.cloudflare.com/d1/platform/release-notes/...) though. They describe removing _at least_ two round trips back to the DC hosting the D1 instance.