3 ms·

> So which cloud provider offers a HARD spend limit? It still baffles me that this isn't the default. Then again, egress/ingress costs also baffle me, why not

> So which cloud provider offers a HARD spend limit? It still baffles me that this isn't the default. Then again, egress/ingress costs also baffle me, why not just offer capped speed and unlimited bandwidth? I doubt that i'll ever willingly use a platform that i pay for myself, which can just decide to charge me bunches of money because of my site/app getting DDoS'ed (assuming not all components are fully protected at the edge or something) or suddenly gaining popularity. In my eyes, such scalability should be opt in (or at least opt out) and by default my apps should just break under the load, much like any self-hosted software would, which may be preferable in some circumstances (e.g. API manages backpressure as best as it can, once request queues fill up, it just gives "503 Service Unavailable" with something like "Retry-After" and any clients/front ends are supposed to show messages to try again later, or automatically retry after a while). To that end, here's a list of providers that give you such fixed resources (in this case VPSes) and that i personally have used in the past: • Time4VPS: https://www.time4vps.com/?affid=5294 https://www.time4vps.com/?affid=5294 (currently hosts most of my sites, hence affiliate link) • Contabo: https://contabo.com/en/ https://contabo.com/en/ • Hetzner: https://www.hetzner.com/cloud https://www.hetzner.com/cloud • Scaleway: https://www.scaleway.com/en/elements/ https://www.scaleway.com/en/elements/ • Vultr: https://www.vultr.com/products/cloud-compute/ https://www.vultr.com/products/cloud-compute/ • DigitalOcean: https://www.digitalocean.com/products/droplets/ https://www.digitalocean.com/products/droplets/ Of course, some of those also offer managed services, but in my eyes it's far safer to just run MySQL/MariaDB/PostgreSQL/MongoDB/Redis and any other piece of software in Docker (or OCI compatible) containers, as opposed to using something proprietary, even with protocol compatibility. That way, it's also possible to migrate between providers easily, should the need arise.

redwood 5y ago
Would you be comfortable with losing all of your data and going down when the spend limit is hit? That's fundamentally why this is not as simple as you think
- KronisLV 5y ago
  > Would you be comfortable with losing all of your data and going down when the spend limit is hit? This doesn't feel to me like a generous enough statement about all of the finer points. Would i be okay with my servers no longer responding to web requests (or doing so with reduced port speeds with some of the above hosts, e.g. Time4VPS) in case of spending limits being unexpectedly hit? Yes, definitely, since my projects going dark for a bit or slowing down would be preferable to me ending up not being able to pay my rent and having to rely on public outcry about the large bill and the vendor's mood that day towards letting that one slide. Would i be okay with my servers running out of space and no longer writing new data to any database, merely responding with the appropriate errors instead? Yes, definitely, because that probably signals the fact that for some reason lots of space has suddenly started getting used for no reason, something that happened so fast that i didn't even get to plan the appropriate scaling, once Zabbix would warn me about 80% of the disk being full (or any monitoring tool would have an alert set up for this). This would probably also be yet another sanity check. Would i be okay with my servers suddenly ceasing to exist and being wiped from existence, for any reason short of excessive abuse complaints or repeated ToS violations? I most certainly wouldn't want this to happen, yet in my experience it has never actually happened - some of the vendors listed allow you to pre pay for the resources you're about to use (e.g. Time4VPS and Contabo), even offering discounts for longer term reservations much like AWS does, without unexpected charges to you like AWS would do. In contrast, some of the other vendors listed allow you to pay based on hourly usage (Hetzner, Scaleway, Vultr, DigitalOcean, at least IIRC), but also won't have unpredictable pricing spikes because of set limits for the most part (ingress/egress charges may still apply, however, a worrying trend in the industry that muddies the waters), if i pay for a 5$ VPS every month, that's what i can expect to pay most months regardless of usage (assuming 100% uptime). With these factors in mind, my servers being wiped would probably have to happen due to either me misusing those services, or alternatively me failing to pay the bills even with the more predictable pricing structures, which is actually very much like a person's electricity being turned off due to them not paying for it - as unpleasant as that'd be, no surprises there. Personally, i don't subscribe to the belief that managed services with unpredictable billing are the only way to do software nowadays, something that Richard Stallman coined as SaaSS (Services as a Software Substitute) and about which you can read more here: https://www.gnu.org/philosophy/who-does-that-server-really-serve.html https://www.gnu.org/philosophy/who-does-that-server-really-s... Alas, as long VPS uptime with predefined resources is the unit of computation that you're paying for, everything else should be fairly simple from there onwards.

redwood 5y ago

Would you be comfortable with losing all of your data and going down when the spend limit is hit? That's fundamentally why this is not as simple as you think

KronisLV 5y ago
> Would you be comfortable with losing all of your data and going down when the spend limit is hit? This doesn't feel to me like a generous enough statement about all of the finer points. Would i be okay with my servers no longer responding to web requests (or doing so with reduced port speeds with some of the above hosts, e.g. Time4VPS) in case of spending limits being unexpectedly hit? Yes, definitely, since my projects going dark for a bit or slowing down would be preferable to me ending up not being able to pay my rent and having to rely on public outcry about the large bill and the vendor's mood that day towards letting that one slide. Would i be okay with my servers running out of space and no longer writing new data to any database, merely responding with the appropriate errors instead? Yes, definitely, because that probably signals the fact that for some reason lots of space has suddenly started getting used for no reason, something that happened so fast that i didn't even get to plan the appropriate scaling, once Zabbix would warn me about 80% of the disk being full (or any monitoring tool would have an alert set up for this). This would probably also be yet another sanity check. Would i be okay with my servers suddenly ceasing to exist and being wiped from existence, for any reason short of excessive abuse complaints or repeated ToS violations? I most certainly wouldn't want this to happen, yet in my experience it has never actually happened - some of the vendors listed allow you to pre pay for the resources you're about to use (e.g. Time4VPS and Contabo), even offering discounts for longer term reservations much like AWS does, without unexpected charges to you like AWS would do. In contrast, some of the other vendors listed allow you to pay based on hourly usage (Hetzner, Scaleway, Vultr, DigitalOcean, at least IIRC), but also won't have unpredictable pricing spikes because of set limits for the most part (ingress/egress charges may still apply, however, a worrying trend in the industry that muddies the waters), if i pay for a 5$ VPS every month, that's what i can expect to pay most months regardless of usage (assuming 100% uptime). With these factors in mind, my servers being wiped would probably have to happen due to either me misusing those services, or alternatively me failing to pay the bills even with the more predictable pricing structures, which is actually very much like a person's electricity being turned off due to them not paying for it - as unpleasant as that'd be, no surprises there. Personally, i don't subscribe to the belief that managed services with unpredictable billing are the only way to do software nowadays, something that Richard Stallman coined as SaaSS (Services as a Software Substitute) and about which you can read more here: https://www.gnu.org/philosophy/who-does-that-server-really-serve.html https://www.gnu.org/philosophy/who-does-that-server-really-s... Alas, as long VPS uptime with predefined resources is the unit of computation that you're paying for, everything else should be fairly simple from there onwards.