Most decentralized compute marketplaces offer only basic Docker containers that share the underlying host machine's Linux kernel. If your enterprise application requires custom kernel modules, eBPF network telemetry, WireGuard mesh peering, or nested Docker daemons, shared containers fail immediately.
Furthermore, when a standard Docker container is paused, ephemeral disk layers are often discarded, forcing developers to re-install packages and re-configure environments from scratch.
NovaPod bridges this gap by offering two distinct compute paradigms: ultra-lightweight Docker GPU Pods (for sub-15s interactive training) and full Root KVM Virtual Machines.
With a NovaPod KVM instance, developers receive dedicated hardware PCI-passthrough, custom BIOS/UEFI boot options, dedicated static public IPv4 addresses, and persistent NVMe root disks that remain intact when machines are paused.
np launch \
--gpu H100_SXM \
--vm \
--image novapod/kvm:ubuntu-22.04 \
--disk 200GB \
--static-ip