Although it’s not as optimal as my preferred platform of B2, Amazon’s S3 should be a considerable savings over using local storage on DigitalOcean. It’s also supported pretty extensively in Discourse, unlike B2.
I’m currently paying the following:
| Resource | Cost/Month |
|---|---|
| Droplet (2 CPU, 2GiB RAM, 60GiB Storage) | $15.00 |
| Droplet Backups | $3.00 |
| Block Storage (5 Days of rolling backups, 75GiB) | $7.50 |
| Total | $25.50 |
It’s worth mentioning that we don’t actually have 75GiB worth of backups. Each backup is 6.7GiB, and there are five of them, so that’s 33.5GiB. The process that does the backups first copies everything to a temp folder, so we assume that takes at least 6.7GiB, plus the 6.7GiB for the archive that’s created from that temp folder, so that’s actually 46.9GiB, which is apparently too low since I was frequently running out of space with a 50GiB volume, which is why I increased it to 75GiB. Also, that leaves room for growth.
None of this is a serious financial burden, but if I can optimize our expenses so that we’re not spending ~33% of our monthly costs on backups, all the better.
Here’s what I think the numbers would look like with S3 for backups (Assuming $0.023/GiB/Month):
| Resource | Cost/Month |
|---|---|
| Droplet (2 CPU, 2GiB RAM, 60GiB Storage) | $15.00 |
| Droplet Backups | $3.00 |
| S3 Backups (5 Days of rolling backups, ~35GiB) | $0.81 |
| Total | $18.81 |
That’s a savings of ~$6.70/Month, or ~$80/Year. The backups would be off-site, too.
I’m not sure how the rolling backups would affect pricing, given that there would be 6 of them for a very brief period each day.
Just for comparison, here’s what the numbers would look like with B2:
| Resource | Cost/Month |
|---|---|
| Droplet (2 CPU, 2GiB RAM, 60GiB Storage) | $15.00 |
| Droplet Backups | $3.00 |
| S3 Backups (5 Days of rolling backups, ~35GiB) | $0.18 |
| Total | $18.18 |
That’s an extra $0.63/Month, or $7.56/Year. Not a huge amount, but worth it if we could go that direction.
The numbers make sense to me. Does anyone have any objections to moving our backups to S3?
The next logical step would be to move uploads (images, etc.) to S3 for live storage. That presents a potentially trickier task since we might have to move existing content to S3, and it doesn’t currently represent a cost savings since we are still within our storage limits on the main disk, but it could be useful down the road as our storage needs expand, and could also provide some sort of performance boost by distributing the network load somewhat.