Aha, I get it now: the GenAI enshittification of cloud services like Slack, StackOverflow, Google, Discord, MSFT, and all the rest is a conspiracy by lonely SysAdmins who miss the old days of running IT equipment in-house and are pushing a return to on-prem 🖥️💽⌨️
another on-premise bare-metal cluster build for 2024!
greater than five but fewer than ten Ampere Altra Q80-30 servers will be combined with Xeon based hosts of mostly-equivalent specs (dual-socket 8280 and E5-2697v4 hosts), 768GB - 1TB of RAM per each, and NICs w/ multiple 10, 25, and 100GbE ports depending on system role. switches are all Arista.
Everyone tells me about „infinity“ scaling and resources in the #cloud. What if I tell you that I can scale pretty well #onprem? I think 1,72TB RAM / Memory is pretty cool…
Guess, I can finally run a single Java app without running oom (hopefully) 😉
Vous le savez, accéder à un #LLM en ligne soulève de nombreux problèmes de dépendance et de confidentialité. La solution est de le faire tourner en local. Mais les obstacles sont nombreux. Le logiciel #onprem en réduit certains : https://www.bortzmeyer.org/onprem-debut.html