KVM and Ganeti give you a virtualization platform you actually own. No per-socket licensing, no feature gating, and no roadmap set by someone else’s sales targets. The trade-off is that nobody hands you an appliance โ€” the cluster has to be designed, and that is the part we do.

Where our experience comes from

We have not only run Ganeti, we have extended it. Work we contributed or built on top of it includes:

  • Custom networking to run VXLAN overlays inside Ganeti, which it had no concept of
  • External storage providers so clusters could use iSCSI back ends
  • IO and bandwidth limits applied per block device, to stop one noisy guest from starving the rest

That work came out of running these clusters day to day, not from a lab. It is the difference between knowing how a platform is supposed to behave and knowing what it does at three in the morning when a node stops responding.

What’s included

  • Cluster design with high availability and automatic failover
  • Storage integration with Ceph RBD or local storage
  • Network configuration with VLANs, bridges, and VXLAN overlays
  • Live migration and disaster recovery procedures
  • Per-guest IO and bandwidth limiting
  • Capacity planning and performance optimization
  • Automation with libvirt and Ganeti management tools

How we work

We size the cluster against your real workload rather than a vendor’s reference architecture, and we prove the failover paths before handover. Live migration and node loss get tested while the cluster is still empty enough that testing them is cheap.