Content
@
https://warpcast.com/~/channel/vercel
0 reply
0 recast
0 reaction
Darryl Yeo 🛠️
@darrylyeo
If I’m understanding correctly, Vercel Fluid is all about persisting / reusing server VMs so a brand new one isn’t spun up for every single request to a server function? Curious why this couldn’t just have been the way Vercel worked from the beginning? @rauchg https://youtu.be/8a4_AAds0vU https://vercel.com/docs/functions/fluid-compute
3 replies
0 recast
8 reactions
rauchg
@rauchg
It’s harder than it sounds. The naive way of doing it can also result in overloading the function instance and degrading performance. The very attractive thing about the 1:1 request to function model was a complete eradication of the “noisy neighbor” problem that e.g.: VPSs suffer from. tl;DR: We focused on quality first then optimized.
1 reply
1 recast
6 reactions
Darryl Yeo 🛠️
@darrylyeo
Makes sense, thanks for the context!
0 reply
0 recast
1 reaction
