We have used MM with now 450 users and Gitlab SSO cost free for years now at work. Can highly recommend it.

Rocket Chat was really never acceptable for us because of 2 main reasons:

• all users on the server are in „1“ team (there is no concept of „sub-servers“ or teams as in Mattermost.
• the cost free push notification limit was always waaaay too low for more than only a couple users.

NAS plus 80 drive LTO9 Library with drive.

AWS Glacier is serval times more expensive FOR ONE YEAR than an LTO library. And if you absolutely can’t afford to get the data back (around 300-400.000$ with 300TB) you can right away not do it at all.

And why would that be? They offer a generous free tier, because they earn their money with companies.

Tailscale as controller based Wireguard VPN is absolutely not comparable to this and way easier to setup and use.

But it can be, with https://github.com/juanfont/headscale .

Apart from that: it is a service with a 3 user, 100 device free tier which supports many other self hosted endeavors.

There is no need to use MAMR for anything up to 24TB, because CMR is possible for those sizes. For 18TB, I would use and have used CMR for years, I have tons of 18TB CMR drives in use.

There are 24TB drives incoming ( https://www.seagate.com/content/dam/seagate/en/content-fragments/products/datasheets/exos-x24/exos-x24-DS2080-2307US-en_US.pdf ), everything below that is available. Only above 24TB MAMR or HAMR will probably be necessary, with 28 TB SMR drives on the horizon.

Rather read the original source, which is way better: https://blocksandfiles.com/2023/11/15/seagate-hamr-drives-come-to-corvault/

What do you even mean with „duplicated stuff“? Again, if your containers are too big, you are doing it wrong.

WTF? You obviously don’t understand Docker at all.

- Docker and Docker images provide the absolute mimimum environment which is necessary to run an application. Containers don’t have reserved resources, so only what is really used is used. A VM has a lot more overhead, as a whole computer plus complete OS are emulated.

- There is not much to deduplicate because there is no redundant storage going on. Only the bare OS essentials plus the app are stored. There are some base OS containers (e.g. Alpine Linux) which are <10 Mbytes in size.

- If containers themselves are “big”, you are doing Docker wrong and store data inside of a container and not externally of the container in volumes or the host filesystem. With the next container pull, that data would be lost.

- no idea what " just not work often due to due to already deduplicated extent stuff" is supposed to mean. That does not even make sense.

Docker is the antithesis of „bloat“.

Just get the original: Seagate Exos. With a MTBF of 2.5 Mio hours, they are twice as „good“ as Ironwolfs.

Screenshot was stolen. Not his system.