So my PostgreSQL deployment processes quite a bit more than 1.2bn records / month (closer to 10bn / month).
What I have to say for AWS (which I use), the back up is why I use AWS. The ease of backing up, rolling back, and connecting additional services, and scaling make it killer.
That being said, I’ve seen around 5x - 10x speed improvement for the same exact cost utilizing bare metal servers. I’ve been using online.net/scaleway and similar services. However the real value save is limited as you need to do a lot more work and often pay more for backing up. Overall, real savings might be 50% cost reduction, with some additional risk. Need more space? Launch a new server, copy, check, switch over, pray. With AWS it’s one click
The real question is what do you want to manage the database or not and how much time you have.
I can attest that pgbarman works great. cough I accidentally nuked a production server once. However thanks to barman the downtime was only tens of minutes.
It works by making a full backup at fixed intervals (say weekly) and then "pretending" to be a replica which makes it receive all the changelogs in realtime. (Basically it receives and archives WAL's). This makes point in time recovery possible.
I've used Barman for physical backups (I used the streaming backup option). Works well, though it's a bit of a slog to get set up the first time (the docs can be confusing in some areas) but once it's up it's great.
Add replication across regions into the mix and things swing even further in favor of AWS. Dealing with cross datacenter replication of MySQL was eating our lunch and preventing us from getting anything else done before we moved to an early version of RDS.
Can you share some information on what were the problems with cross region replications? Does AWS use a replication protocol / transport other than the MySQL's built-in replication?
the real question is also, when AWS becomes the "authorized, secure" source to whatever, while independent network nodes are marked as "unauthorized, untrusted" via some opaque mechanisms
What I have to say for AWS (which I use), the back up is why I use AWS. The ease of backing up, rolling back, and connecting additional services, and scaling make it killer.
That being said, I’ve seen around 5x - 10x speed improvement for the same exact cost utilizing bare metal servers. I’ve been using online.net/scaleway and similar services. However the real value save is limited as you need to do a lot more work and often pay more for backing up. Overall, real savings might be 50% cost reduction, with some additional risk. Need more space? Launch a new server, copy, check, switch over, pray. With AWS it’s one click
The real question is what do you want to manage the database or not and how much time you have.