Message from @porco
Discord ID: 542007644190146595
although I could just add an ssh tunnel
and run it all on ubuntu server or smth
Let me shill you some HPE ProLiant MicroServer
it's more generic than your typical NAS since it's a server, but it's smol and has many bays
also not so expensive
you can obtain them with already built in ram and no disks for about 350 eur
sadly, the newer HPE MicroServers don't have ILO included anymore, ILO was nice
ILO is HP's web based hardware / system management tool
it runs on a firmware level so you can access it even if the system doesn't boot
but the newer "budget" HP servers like the MicroServer don't have it anymore
what's the budget anyway? (without disks)
tfw no dual epyc server
i wonder what situation you would need dual epycs
over just GPU accelerated server
posting screenfetch on /g/
I think Lucene / Elastic etc might actually profit from that
because they need fast cores to operate, and many cores to scale
imagine you're a webhoster, one of these (Let's say about 5k total investment to put the missing hardware inside) will make you be able to sell 128 VPS with 1 Core / 1 GB Ram @ 10 $ / month
=> 30k / year income
question would be if multiple boxes larger cost of footprint but cheaper total price be worth it
probably not cheaper
just a lower initial investment to get started
the $ / core count on EPYC is incredible
And selling 1 HT as 1 Core in a VPS is a standard since Microsoft and Amazon both do it
which means on that machine, you can sell up to 128 HTs
you're barely going to be more profitable than with that kind of set up. Also consider the costs for RHEL, Virtuozzo etc etc
implying they don't oversell cores
no they don't
i mean MS/Amazon don't
only small ISPs overcommit cores
but a lot of the small time ones do
yes sure
small ISPs overcommit cores and ram
budget on total is sub 1000 €. preferably under 700 €.
if you're a big company you don't care as much about $ / core count as you do as core count / watt
k8s makes you allocate "microcores". Basically 1/1000 of a core
so you could have each vm have 500 microcores