←back to thread

Researchers have built server prototypes that re-use old components

(spectrum.ieee.org)
45 points rbanffy | 2 comments | 15 Nov 24 11:32 UTC | HN request time: 0s | source
Show context
jdoss ◴[20 Nov 24 07:41 UTC] No.42191598[source]▶
I have a rack in a datacenter with mostly 5 year old Supermicro servers. We bought them all off of Ebay for no more than $400 each. They work great and we have more compute and bandwidth for our workloads for less than $1000 a month. If we used one of the could providers it would be many thousands of dollars per month. I understand not everyone has the skills to run their own rack but the value of doing so is totally worth it.
replies(3): >>42192665 #>>42192958 #>>42196068 #
1. pdimitar ◴[20 Nov 24 10:49 UTC] No.42192665[source]▶
I for one never learned to do it.

IMO there's a big historical and archival value in the idea to start cataloguing such knowledge.

replies(1): >>42193972 #
2. mxuribe ◴[20 Nov 24 13:58 UTC] No.42193972[source]▶
Agreed! And there's this space in the middle there...Because at the dayjob/enterprise side of things there's the concepts of hypervisor vendors, CoLos (co-locating servers in data centers), and full-blown enterprise on-premise...and of course other end of spectrum is homelabbing/that is self-hosting stuff at home either on consumer or old server grade computers...but what if i want to have a single server for my own self-hosting use but in a tiny data center/colo, but which is not outlandish, and i would maintain it myself? I dont know what we'd call that space: mini-CoLo, or self-data centering, or remote-homelabbing? but, i'm sure there's a gap in knowledge there, and would be great to learn more. :-)