Human error is the biggest factor.

You edit a file, do you directly copy it over? Probably not, so when is it time to copy over? Maybe you remember maybe you don’t.

Ohh you Deleted a File? Did you delete it on the other one aswell? No, well now we get into the rabbit hole of dependencies.

Sure you may be able to automate it via Rsync. But how often? once a day? once a month? what happens if a Drive writes faulty data and does not tell you? Well now you copied it over in the hopes it is not broken.

You want to retrive a file because you accidentially deleted it, well that file has been sitting there a long time, its corrupted now. But it has never told you it is and you have no redundancy to correct errors.

Theres many more reasons just copying it is not a good Idea those are just some examples. Theres also Raid soulutions to accommodate mixed drive sizes like Unraid does. And yes it costs money but with Snapraid and Merger FS you can do the same thing for free.

Sure it maybe more effort but the question should be how much is your data worth and are you able to lose it? And what does it cost you to get it back again.

86W so about 20Euro/Month

Not to be rude or anything, but External RAIDs individual to the user is not really a solid soulution. It may work for 1-2 People working on one project at a time. But it just does not scale. What if someone needs to acces files of that project? they move the raid or plug their laptop on a differen workspace? Not really a great soulution IMO.

Like you say in the last part having a NAS with maybe a bit of room to grow sso 100TB might be the best option that way everyone can access the data and work accross projects. And more importantly it would offer work from a different place in the office or even work from home.

Yea with tape the compressed nr are very missleading. Thats a best case scenario where the files compress 2:1 with TAR+gzip which it literallly never does. Bestcase I have seen was 1.2:1 on a folder consisting of config files. Basically nothing nowdays is compressable you will interact with, except textfiles depending on format. So its best to always asume the raw space as the space you get

From what it sounds you want a NAS and Tape Archive.

So get a device which holds your working Projects, you mentioned arount 20-40TB which is no problem nowdays. Can be done for under 1k with of the shelf stuff.

And Tape backup for stuff you dont need regularly. Maybe chose an older generation of LTO I would look for something that can hold about 1 Project per Tape or the likes of it. LTO5 is pretty cheap used, ca be had for 500 Bucks but is only 1.5TB per tape.

Disclaimer, with LTO never look at the compressed NR, its for compressable data only which video is not. Thus with LTO9 you will only get 18TB

Why not go with a constant quality setting in the encoder instead of bitrate and filesize limitations?.

Thats what I do I feel CQ of 18 is indistiguishable from the original.

Bonus TIP when transcoding for static Use like storing and filesize reduction. always use a CPU never a GPU if you are after smallest filesize.

Well ddns is just dns which get auto updated via helper script which you can do yourself on your own domain aswell. So yes Point still stands. And a domain which is yours might still be nicer than a bottom of the barrle subdomain. Of which you often also don’t get unlimited

SSD is not archival storage. The NAND needs Power or it will lose its charge over time. Maybe not in one to 2 Years but maybe in 5-10.

If its active yes hdds are still cheaper unless you need less then 10tb

Yes, because how will you acces wireguard? Whats your endpoint?

I mean its all fine if you have a static IP which never changes but that usually is not the case anymore. So you need a domainname to update. Also vpn.youlab.tld is esier to remember than 131.234.142.83

WIth any kind of advanced software you will never be getting drive speed out of a soulution.

Raw nvmes can hit 800k iops. Add XFS and you may be able to get that.

with MDADM you get like 200-300k

ZFS shrink it to 20-50k

Anything network be glad if you get 5-20k

The more software is involved the worse performance gets especially for IO. Sequentials often scale better or even linearly, but IO is a PITA

I mean this is too litle info. What IP exactly? one v6? one Legacy IP? also HDD storage? bruh…Whatabout SLAs? RAM CPU?

Datasecurity? Redundancy?

Even if its a friend I would avoid it.

SAS is not networking. If you expect to connect 2 systems to another it wont work.

If its just for disks then yes.

I monitor/log it with a smart plug, or UPS.

At my home I have a shelly plug s Which is just a wallplug insert. That monitors consumption and can give it to you via MQTT, I port it to home assistant and monitor it. Usually my small HP mini node with a couple switches and Router is about 80w.

At my parrents I have a Shelly 1PM Plus, which is integrated in the path to the UPS. It monitores everything like the PoE Switch APs, Server etc as everything is connected to it thats IT. Its about 4.5kwh/day so about 220-240W. That gets also monitored via Home Assistant and MQTT. and 4.5khw/d are bout 2.5USD/day for me. So about 920USD/Year.

So Defo more expensive than a VPS would be. But also more custom and more of an Experience to gather.

My advice if you don’t want your house burning down. Dont fiddle with it, thow it away, or make an RMA claim.

It not like its a 100 Buck product. Saftey goes first here.

It all depends. First you dont have to get your own hardware. Sure some of us get stuff from work for free or have an old PC but even a 100 Buck PC is 5 Months with linode for example. Secondly, Linode is not the only option for example Netcup offers a VPS with 4c 16gb for 10-15 a month. So that seems a bit more fair IMO.

ANd theres other reasons. Like: You want to selfhost. But don’t get a public IP. Or your internet service being so bad that its not worth it. Which are all issues you don’t have with a VPS. Further Uptime. At home to keep a good uptime you would probably want to invest in a UPS and other hardware to eliviate problems. which further costs money which you could invest into a VPS.

And the Killer for me is Power Consumption. At 40-50ct/kwh running a PC of 60w costs me 300+ a Year, whereas a VPS at 15-20 costs me 240. And I get the benefit of lower latency and better bandwith and not needing to pay for extra Internet service.

So it depends. I do still selfhost stuff at home, why? not because Its cheaper, it actually isn’t but rather to gain experience with the hardware etc. But i do see that many people may not want to deal with that. So a VPS is defo a viable option.

Soo you are saying you draw a random ass conclusion from an issue which would not exist if you would have made the right hardware choise? got it. Well done.

Experience. But it all depends. TDP gives a hint in some generations example epyc cpus TDP is relaivly acurate wheras the 35w tdp of intel second gen means litteraly anything between 10 and 60w.

But for consumer systems with an i or r5 cpu and a mid tier gpu 500w is enough. For a server without a gpu and similar cpu 300W will suffice.

dual socket systems of that cpu class maybe 400-500.

With more enterprise stuff 500-700 or with epyc maybe encroaching 1000w with some addin cards.

It depends on maany other variables like Drives attached, PCIe devices used etc.

It’s never enough

Further be aware that PCIe devices also need to support sleepstates. So if you add an old enterprise card liek 10g to your server/machine, then you might idle at twice or more times the power it would otherwise do as that device prevents lower sleepstates.

so the cheapest 10g card or hba is not the best in the long run

Lol that same bill would slap me with a 1.3kEuro/A bill lol at 40ct/kwh

I am paying the same 270 for my 60W lab.

Great talking points tho. Power is always somethign to considder whenever I see someone getting a big enterprise server.

Only question I have is: The fuck? Why is there so many standards and different outlet types?

As someone Living in Germany we have exactly one for basically everything.

With the only exceptions being higher current industrial uses in single and 3 phase. and they are also the same among themself only in 16-32 and for 3phase in 64 and 128A variants.

That picture is madness.