Why Local Disk Stops Working
It works on your machine — it breaks the moment you redeploy or scale
Writing an uploaded file straight to the server's local filesystem is the simplest thing that could possibly work, and it genuinely does — right up until the conditions every real production app eventually hits.
Where local disk storage breaks down
- Redeploys wipe it — Most modern hosting platforms (Vercel, most container platforms, many PaaS providers) run on ephemeral filesystems — anything written to disk during a request is gone on the next deploy, sometimes on the next request
- It does not survive scaling past one server — A file uploaded to server A is invisible to server B — the moment you run more than one instance, half your users get 404s depending on which instance handles the request
- There is no real durability guarantee — A single disk failure can take every file with it — no replication, no redundancy, unless you build that yourself
- Backups are your problem entirely — A managed storage provider replicates data across multiple devices and often multiple facilities by default; a server disk does not, unless you set that up
The self-storage unit analogy
Object storage is a self-storage facility, not a closet in your apartment
A closet in your apartment is fine until you move — everything in it has to come with you, and if your apartment floods, so does the closet. A self-storage facility is built specifically to hold things long-term, independent of where you happen to be living, with its own security and redundancy. Object storage is the self-storage facility for your files — it exists independently of any one server, survives that server being replaced, and is built specifically for durability.
What real object storage gives you instead
- Files survive redeploys and scale across servers automatically — Every instance of your app reads and writes to the same bucket, regardless of how many instances are running
- Durability guarantees measured in nines — Major providers advertise 99.999999999% ("eleven nines") annual durability — your data is replicated across multiple devices by default, not something you configure yourself
- A predictable, usage-based bill instead of a disk you have to size and monitor — You pay for what you store and transfer, not for provisioning a disk large enough for next year's growth today
Try this
Check how your current project handles file uploads. If the answer is "writes to a folder on the server" or "writes to /tmp", that is the exact gap this course closes — note which upload flow it is, you will replace it with real object storage by Lesson 3.