My question: Is out there any other PostreSQL managed service faster than Azure and as simple to set up?
If that is not the case then I'm in trouble. I write and benchmark applications on my own laptop and always assume that if my low-end laptop is fast enough then whatever production-grade infrastructure we deploy it on will be on-par or faster. Seems like this assumption breaks down when it comes to Azure.
We are thinking in upgrade it to a better plan, but we are not sure because we are not sure that it will solve our problems.
I think that comparing to your own laptop is unfair because there is no connectino needed, everything is on the same CPU / RAM, etc. Remember when you connect to a external DB informatin has to travel from one point to another.
> I think that comparing to your own laptop is unfair because there is no connectino needed, everything is on the same CPU / RAM, etc. Remember when you connect to a external DB informatin has to travel from one point to another.
Yes indeed, but in my case I was talking about DB query performance where the network plays a minimal part. If a query is near-instant on my laptop I expect it to be the same (or better) on production infra, but with Azure's Basic tier I noticed the opposite. I agree that if you're firing off lots of queries then network latency starts to add up, but in my case I haven't even reached that point - network latency is insignificant compared to the actual delays I'm seeing on the individual queries themselves.
We're lucky to have experience with Postgres, performing backups (of the various types) and restoring DBs on production environments. We don't have things automated, but that would be next.
I think startups are hard to do it in-house, not for a lack of expertise (or willingness to get it), but the fact that there are already so many things to handle.
Can you share your experience with using soyoustart?
The experience is mostly solid, though they don’t provide any kind of console access and I once got locked out after putting a machine to sleep from the command line (and had to open a ticket for someone to reboot it manually).
The "problem" is that this is not a virtual machine and most assumptions or ways of thinking that work for AWS or similar don't work here. For example, you can't just upload an image or move data volumes around between machines. The storage volume is a physical disk attached to the machine, if you need to copy them you need to actually boot the machine, SSH into it and do your thing. Similarly, those disks can break down and you need to plan ahead (they provide 2 disks per machine and expect you to use a software RAID 1 or have a backup/restore strategy that can tolerate those disks failing).
This is where building your own layer on top of it (whether Kubernetes or VMWare or similar) makes sense and will make your life easier and allow you to use most tools we take for granted when it comes to VM providers - you're essentially building your own AWS.
This is a whole other league compared to Linode - the former provides VM instances with varying levels of performance - the one we’re talking about is bare-metal and you are guaranteed to have the performance of whatever hardware specs you chose. Bandwidth is also unmetered so you can saturate the pipe 24/7 if you wanted to. Very cost-effective for serving media, though every machine only comes with 250Mbps, so maybe it's better to have an origin server on SYS, and lots of downstream instances on Kimsufi which serve as a downstream cache for the origin - they are dirt-cheap and come with 100Mbps bandwidth each (unmetered as well).