My 3-2-1 Backup Strategy: Why I Swapped MEGA S4 for Scaleway S3

A 3-2-1 backup rule is the industry standard for data survival, dictated by the logic that any single point of failure is a guaranteed disaster waiting to happen. It requires keeping three copies of your data on two different media types, with one copy stored offsite to protect against local catastrophes like fire or theft. By decoupling your backups from your physical location, you ensure that even if your home lab goes dark, your data remains accessible and intact.

I had originally been using Backblaze for offsite backups as it proved reliable, affordable, and straightforward. Despite being an American company, it served my needs well for the offsite leg of my 3-2-1 backup strategy. Every night, my Proxmox Backup Server would sync my local backups to both my NAS and Backblaze’s S3 bucket.

Then I discovered that MEGA, which I was already using for my daily cloud drive, had recently launched their S4 storage solution, compatible with S3. For me, it was a no-brainer. Since I was already paying for the subscription and had plenty of room left on my 3TB plan, I figured I might as well leverage that for offsite backups too.

The Problem: Slow, Sluggish Web UI

But I soon ran into a frustrating issue. I don’t access the MEGA web interface often, Nextcloud is my Primary cloud storage, self-hosted of course with MEGA being the offsite backup for my personal files so I didn’t notice the problem at first as the desktop sync app was still working flawlessly.

Then, a few days ago I tried to open it, and it was painfully slow. decrypting my data, which normally takes seconds, was still ongoing after 5 minutes and sometimes failing to load all together. I tried different browsers, different computers, nothing helped.

When I finally got into the interface, the root cause was obvious. MEGA had embedded a link to my S3 bucket within the web GUI. Great for quick access, but disastrous for performance when your backup contains tens of thousands of tiny chunks. The GUI had to decrypt and load every single one of those chunks every time I visited the page. The result? The interface was sluggish, unresponsive, and frankly, unusable.

MEGA’s architecture is primarily designed as a cloud drive, it isn’t optimised for a PBS workload. Its web UI does a full account metadata scan and decryption every time, one of the reasons I chose MEGA in the first place, if I lose my access credentials, MEGA cannot access my data or reset my password. For small, encrypted chunks, this becomes a major bottleneck. It is the digital equivalent of trying to read a library where every single page is in a separate locked box, the sheer overhead of unlocking each one makes the actual content irrelevant.

The Solution: Moving to Scaleway

I finally decided to switch. I chose Scaleway over Backblaze this time for a few key reasons: they are EU-based, meaning my data is protected under stronger privacy laws, and their prices are competitive. Setting up was straightforward. I created a new account and migrated my data over.

Unlike MEGA, Scaleway treats the bucket purely as object storage. No drive overlay, no metadata chaos. This means fewer bottlenecks, faster performance, and a more reliable experience.

I keep only the critical backups offsite, about 100GB. Just enough to ensure my most important data is protected without breaking the bank.

How I configured it

• Region: I selected the EU region (e.g. fr‑par) for data sovereignty.
• Endpoint: Used the specific S3 endpoint (s3.fr‑par.scw.cloud).
• Remote setup: Configured PBS to connect directly to Scaleway’s S3 and enabled the local metadata cache to reduce API calls.
• Transfer mode: Transfer Last = 1 and Remove Vanished enabled to keep the remote tidy and costs low.
• Garbage collection: Run PBS GC after each sync to remove unreferenced chunks, saving space and avoiding waste.

Privacy and Security

Since PBS encrypts data locally before uploading, Scaleway only ever sees ciphertext. I hold the encryption keys, stored securely off-host.

This simple, efficient, and privacy-conscious architecture adheres to the classic 3-2-1 rule while respecting my hardware limits and budget.

Final Thoughts

Automated sync jobs run silently on my dedicated Proxmox Backup Server at 04:00 (local) and 05:00 (cloud), handling backups smoothly in the background. This setup proves that you don’t need big tech giants to own your data. With a bit of planning, you can keep it private, reliable, and affordable.

Owning your infrastructure isn’t just about control, it’s about peace of mind.