Developers - How much Bandwidth is reasonable

msatt

Working on some great new features for FOSSVPS front end.
But before I release anything I would like your opinion.
FOSSVPS is for developers of Open Source Software. It is not for 'hosting'. So what would be a reasonable amount of data transfer in / out that should be allocated per client per month. Bear in mind that projects would need down / uploading.
Genuine feedback appreciated

rpqu

Again? @msatt, I would keep it different per IP/domain/AS.
For example, github/gitlab/codeberg should be pretty clear. I know there's workaround to use what's it called? space something? Cloud drives should have different limits. OS, C*AN, other package management could be pretty much unrestricted. But, if you see network activities detached from disk activities, that means it's being used as relay. However, that depends on how much you could tolerate.

---

10GB is reasonable amount of traffic. Unless it's network related

MallocVoidstar

Google gives me like 10GB free per month, I haven't touched my Google Cloud free VPS in years and now they're charging me $0.01/mo in bandwidth overages. So 11GB.

somik

Summary: keep it somewhere around 100 to 500 GB per month (ingress + egress)

.

Details:

Profile	Ingress (Download)	Egress (Upload)	Use-Case Breakdown (VPS-only)
Light VPS usage	5–30 GB/month	10–80 GB/month	Ingress: SSH git pulls (1–5 GB), small dependency installs (1–10 GB), occasional container pulls (0–10 GB), light DB backups/restores (0–5 GB) Egress: small web/API traffic (5–50 GB), bots/webhooks (1–10 GB), file serving (1–20 GB), logs/monitoring export (1–10 GB)
Moderate VPS usage (typical SaaS dev)	50–300 GB/month	100–800 GB/month	Ingress: CI/CD deployments (10–50 GB), Docker image pulls (10–100 GB), server-side package installs (10–50 GB), DB backups/migrations (10–100 GB), log ingestion pipelines (5–50 GB) Egress: API traffic (50–400 GB), web backend responses (20–200 GB), file/media downloads (10–200 GB), webhook/integration traffic (10–50 GB), logs/analytics exports (10–50 GB)
Heavy VPS usage (DevOps / AI / infra)	0.5–3 TB/month	1–10 TB/month	Ingress: large container images (50–300 GB), ML models/datasets (100–1000 GB), CI build artifacts (50–300 GB), DB replication imports (50–500 GB), cluster bootstrap/sync (50–200 GB) Egress: high API traffic (500 GB–5 TB), image/video/AI outputs (200 GB–3 TB), file/media serving (100 GB–2 TB), distributed system traffic (100 GB–1 TB), logs/metrics streaming (50–200 GB)

Note: Edited with AI for easier readability

AnthonySmith

I can't imagine anyone should genuinely ever exceed 50GB and if they do should they really be using FOSS to begin with?

50GB should be the genuine use case 'i don't need to worry " number and I guess if someone is an exception they can ask for a one time bump

msatt

Excellent feedback - thanks everyone.
So perhaps something like 40GB / month with a 20GB bimonthly auto boost if a client goes over.
So I scale based on an average of 50GB per client per month.

rpqu

@msatt said:
Excellent feedback - thanks everyone.
So perhaps something like 40GB / month with a 20GB bimonthly auto boost if a client goes over.
So I scale based on an average of 50GB per client per month.

40GB should be sufficient unless you're pulling 5GB docker image at each commit. But, at 5GB, I think that's excessive and signals bad dockerfile.

But, what's the story behind it?

msatt

@rpqu I am hopefully introducing a new frontend which shows b/w usage plus max allowed. I don't want to encourage over usage or limit client functionality - I know - don't show it, but I would rather everything is clear and open.
There is also the overall node usage, so FOSSVPS do not abuse our generous donors - @Alexhost, @Hosteroid and @st-hostings.

edit - typo

rpqu

Okay. It thought someone had used excessive traffic.

somik

@rpqu said:

@msatt said:
Excellent feedback - thanks everyone.
So perhaps something like 40GB / month with a 20GB bimonthly auto boost if a client goes over.
So I scale based on an average of 50GB per client per month.

40GB should be sufficient unless you're pulling 5GB docker image at each commit. But, at 5GB, I think that's excessive and signals bad dockerfile.

But, what's the story behind it?

Docker pulls gets cached. So next pull would use cache and not get from docker hub. Only pushes would count. I have docker images that are 2+gb but those are rare and are only big cause of updates and latest java sdk. Those are not used during deployment, only during builds.