DocumentationAbout e-INFRA CZ OpenStack IaaS Multisite Cloud Oliver
e-INFRA CZ OpenStack IaaS Multisite Cloud Oliver
IaaS OpenStack cloud providing cloud services for following organizations:
Unlike the G2 Brno and G2 Ostrava clouds, which are two separate OpenStack deployments, the Oliver cloud is
a single multi-site OpenStack cloud (one identity, one API/Horizon endpoint) spanning both geographical locations.
Compute capacity is split into two host aggregates / availability zones — brno and ostrava — so you can choose
where your instances run.
Hardware
The cloud runs on 24 active nodes (Kubernetes cluster size verified by the cloud team, August 2026) in two geographical locations:
Brno site (Faculty of Informatics, Masaryk University, datacenter A510) — 17 nodes:
| Node group | Count | Role |
|---|---|---|
hdn-* | 5 | compute hypervisors (e1.*, c2.* flavors), 64 physical cores (128 threads) per node |
hdp-* | 6 | compute hypervisors (c3.* flavors), 128 physical cores (256 threads) per node |
cp-osd-* | 6 | dedicated Ceph storage nodes (NVMe, encrypted devices) |
Ostrava site (IT4Innovations datacenter) — 10 nodes:
| Node group | Count | Role |
|---|---|---|
o-hda-* | 10 | Gigabyte R183-Z94 nodes, 128 physical cores (256 threads) per node; 5 serve as compute hypervisors (e1.*, c2.*, c3.* flavors), the rest hosts the OpenStack control plane |
The tables list the full roster of machines assigned to the site (27 total). The active node count above reflects what is currently joined in the underlying Kubernetes cluster and changes as hardware is commissioned, serviced or rotated.
Software, cloud versions and components
The Oliver cloud site is built on top of OpenStack, which is a free open standard IaaS cloud computing platform and one of the top 3 most active open source projects in the world. New OpenStack major versions are released twice a year. OpenStack functionality is separated into more than 50 services.
OpenStack 2025.1 (Epoxy) deployed on Kubernetes 1.34 via openstack-helm.
Following OpenStack components are available.
Virtual server drives, volumes and image storage are provided by distributed Ceph clusters at both sites (see Storage below). Tenant and control-plane networking uses OVN on top of OVS (Open vSwitch) with geneve overlay tunnels.
Cloud architecture strictly follows IaC, GitOps approaches to minimize cloud maintenance toil.
Cloud workload
Both general-purpose (e1.*) and compute-oriented (c2.*, c3.*) workloads up to 120 vCPUs / 240 GB RAM per instance
are supported. See List of flavors.
Storage
All storage in the Oliver cloud — instance root disks, additional volumes and images — is provided by distributed Ceph clusters. There is no ephemeral (local hypervisor) storage; instance data always resides in Ceph.
Two block storage backends are available:
| Backend | Location | Media | When to use |
|---|---|---|---|
| Fast Ceph pool | Brno | NVMe, triple replication, encrypted devices | default backend; choose for I/O performance-sensitive workloads |
| Capacity Ceph pool | Ostrava | rotational disks (HDD) | large-capacity or less performance-demanding storage |
Volume types are named generically after the backend. Run openstack volume type list or check the volume type
selection in Horizon to see the exact type names currently available in your project.
VM image backup is work in progress
Automated backup of VM images is currently being rolled out on the Oliver cloud. Until the service is available, you are responsible for backing up your own data — see the volume backup guide and our terms of service. See also the site FAQ.
