S3 Object Storage

I try to use a CDN to solve any scaling issue I can -- vs. databases and memcached layers.

We use BunnyCDN and Backblaze B2.

These both a surprisingly fast TTL that lets you use your cheap CDN a lot more like a slow database than a fast CDN.

They're cheap, fast, and awesome. And I can remember each bit of downtime they've had and what caused it for the past 2 years.

So, we store objects in both to give us a failover option (that we can enable via DNS very quickly).

I store images, JSON docs, and a small amount of video.

These are not great solutions for video though (you have to build a lot of tooling on top of them).

There's two reasons why you'd use AWS or Google instead of these:

1. You care a lot about nine nines of availability. Neither of these are perfect, but they will have better uptime than BunnyCDN or Backblaze B2.

2. Your objects would do a lot of ingress or egress from other instances in AWS or GC. The total cost might be lower for you if you have to connect to a lot of instances in either of these cloud services.

No one is measuring storage costs. The costs are all transfer for these, and if you don't have enough transfer to track, then you're never really going to have to worry about the cost of these services.

Transfer to odd regions (China, Africa) end up costing more than they should in all providers.

DigitalOcean spaces isn't great. I don't like Wasabi's business model, although it worked fine when I tried it, they obviously price below cost.

1- S3.
2- Yes
3- I don't, but I would due to transfer costs that are insanely high at AWS. I'd go to DigitalOcean, which is far cheaper.
4- Audio, video, images, some JS, and CSS.
5- Write once per deploy, read a lot - static website
6- Transfer costs
7- Latency/BW is a concern, as it affects the TTFB of my page.

Several different uses for different projects.

Project 1 .
User avatar images, user asset images, other images uploaded to application. Access pattern. Write once, read thousands of times.

Project 2.
Document storage. Dropbox is often blocked by corporate firewalls, so we utilize s3 to share marketing materials, presentations etc.. as part of marketing and sales processes. Write once, read thousands of times.

Project 3.
Long-term DB backup storage. Write once., never or rarely read.

In additional to AWS, I've also used the GCP equivalent, but only for DB backups. AWS is generally the default for me because the libraries are already mature and I'm familiar with them.

Hi Nick,
I use AWS and Digital Ocean completely to run my web Scraping API - www.scrapingproject.com. AWS Lambda, S3, etc.

1. S3
2. No, I use Digital Ocean to host my website.
3. I found it easy to deploy Rails App in DO.
4. Logs
5. For Analysis.

I hope it helps.