Conversation

Notices

1. Embed this notice
   Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Saturday, 01-Mar-2025 16:07:39 JST Haelwenn /элвэн/ :triskell:

   • Cinnamon
   @q66 @cinnamon Plus there's issues beyond mere CPU speed, like latency and bandwidth doesn't follows Moore's law.
   In fact network latency is pretty much stuck in place as it's quite close to physical limits.
   • Embed this notice
     Cinnamon (cinnamon@mk.absturztau.be)'s status on Sunday, 02-Mar-2025 02:59:35 JST Cinnamon

     in reply to

     @lanodan@queer.hacktivis.me @q66@gts.q66.moe Then I see network latency as something (even more) against scalable designs (which communicate nodes using a network). A design centralised in a beefy single server seems better. Am I right?

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Sunday, 02-Mar-2025 03:10:35 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • Cinnamon
     @cinnamon @q66 Well, it depends on what you're serving and how it's done.
     Like serving rarely written to files (or even write-once) will probably still be scalable for decades, specially when it's putting nodes around the world as that reduces latency.
     And it can help getting redundancy.