Making microvms fast, persistence and cheap on GKE with Kata and CH Kosem cut Linux 6.6 kernel source unpacking from 126–129 seconds over virtiofs to 7.6–8.8 seconds by attaching Hyperdisk volumes as block devices over virtio-blk instead of sharing them via virtiofs, according to the company's engineering writeup on running Kata Containers and Cloud Hypervisor microVMs on GKE nodes with nested virtualization. Adding mke2fs flags (nodiscard, lazy_itable_init, lazy_journal_init, -m 0) cut formatting from 101 seconds to under a second, bringing cold boot on a fresh disk to 20 seconds and restart to 4.7 seconds, while stat on 20,000 files got 45× faster. The writeup also reports that the container root filesystem still runs over virtiofs, where one iteration of Terminal-Bench's cobol-modernization test loop took 106 ms in the microVM versus 9 ms on a laptop. Making microVMs fast for coding agents. What we learned running Kata and Cloud Hypervisor on GKE Coding agents don't only edit files. They install system packages, run test suites and start Postgres, Redis and BuildKit with docker compose . That needs a real Linux kernel, and running untrusted code for many customers needs a hardware boundary around it. So every kosem environment is a microVM: Kata Containers on Cloud Hypervisor, on GKE nodes with nested virtualization.