Ordered a pair of clogs from @Flipkart on Jul 12. ETA was Jul 14, then delayed to 16, 18, 19, and now 20! Every support call blames @shadowfax_in and says "issue escalated," but nothing changes. No delivery attempt, just copy-paste replies. What is going on? Do you even care?
@NammaBESCOM Honestly, you guys are a genius-level anti-theft deterrent. Can’t steal electricity if there’s no electricity to steal, right? it's the most effective security system I've ever seen
@NammaBESCOM No power in Marathahalli, ckb layout.
Whats happening??? every day there's power outages in working hours , are we living in farm or what ??
What is the timeline please be clear ?
That’s why this approach works well for serverless platforms.
You get something close to container speed, but with a real virtualization boundary.
Enough isolation to run multi-tenant workloads safely.
Why not full-featured VMs?
Because they optimize for a different lifecycle:
- hotplug devices
- live migration
- long-lived instances
- complex device models
All of that adds flexibility, but also latency and state.
For short-lived compute (functions, CI runners), you want the opposite:
predictable boot
minimal state
fast teardown
tight isolation
Small VM shapes fit that better.
The machine model is intentionally narrow:
1. minimal virtio devices
2. no broad hardware emulation
3. fixed boot path
4. separate guest kernel
That keeps startup fast and reduces the attack surface.
Under the hood it still looks like a VM lifecycle:
open hypervisor → create VM → map guest memory → create vCPUs → load guest kernel → run → teardown
The difference is how little is exposed.
“Serverless” didn’t remove servers.
It changed the unit of compute.
The shift was making VMs small and fast enough to use per request, per job, per workflow.
A thread 🧵
#vm#serverless#linux#cloud#platform