OpenStack Expertise: Deployment, Integration and Operations

OpenStack is the leading open-source foundation for building an IaaS private cloud. LaMeDuSe Cloud designs, deploys and runs OpenStack clouds, on our European infrastructure or on your own hardware, with Ceph for storage and 24/7 monitoring.

The problem you're facing

OpenStack is powerful but demanding. A dozen interdependent services, twice-yearly releases, complex software-defined networking and storage that has to keep up: many projects start with enthusiasm and run out of steam in production, for lack of a team able to maintain the platform over time.

Our answer

We take on OpenStack end to end: an architecture sized for your use, automated and reproducible deployment, integration with Ceph, your directory and your tools, then day-to-day operations. Your teams enjoy a self-service cloud; our engineers keep it running.

Why OpenStack for a private cloud?

OpenStack turns a set of servers, disks and network equipment into a real cloud: users who create their own virtual machines, networks and volumes, projects isolated from each other with their own quotas, and a complete API driven with Terraform, Ansible or standard SDKs. It's the public cloud experience on infrastructure dedicated to you.

It's also a choice of independence. OpenStack is free software, governed by an open foundation and used by telecom operators, universities, public administrations and hosting providers worldwide. No vendor can unilaterally change its licence terms: an argument that has gained a lot of weight since the pricing changes of some proprietary virtualisation products.

Architecture: OpenStack's core services

An OpenStack cloud is made of specialised services that talk to each other through APIs and a message queue. We deploy and integrate the core services, then additional ones as needed (Octavia for load balancing, Designate for DNS, Barbican for secrets, Magnum for Kubernetes, Swift or Ceph RGW for object storage).

Physically, the architecture separates control nodes, which host the APIs, databases and message queue in high availability (at least three nodes), compute nodes that run the virtual machines, network nodes that handle routing, and a Ceph storage cluster.

  • Keystone: identity, authentication and service catalogue
  • Nova: virtual machine lifecycle
  • Neutron: software-defined networking
  • Cinder: block storage volumes
  • Glance: system images
  • Horizon: self-service web console

Nova: compute

Nova manages the instance lifecycle: creation, resizing, live migration, shutdown and deletion. Its scheduler picks the right compute node for each virtual machine based on available resources and the constraints you define: CPU and memory flavors, affinity or anti-affinity rules to separate instances of the same application, and host aggregates to reserve specific hardware (GPUs, high-frequency CPUs, local NVMe disks).

With shared Ceph storage, live migration of virtual machines becomes routine: we can drain a compute node to update it or replace a part without interrupting your services.

Neutron: networking

Neutron provides self-service networking: each project creates its own private networks, subnets, virtual routers and security groups (the equivalent of a firewall attached to each instance). Floating IPs expose to the Internet only the services that need it.

For the implementation we rely on Open vSwitch and OVN, with overlay networks (VXLAN or Geneve) that isolate projects from each other, and provider networks when instances need direct access to your existing VLANs. The physical network design (redundant links, separation of management, storage and data traffic) is often what makes the difference between a stable and a fragile cloud: we address it from the architecture stage.

Cinder and Glance: storage

Cinder provides block volumes that are attached to instances, extended, backed up and snapshotted. Glance stores the operating system images that virtual machines boot from.

We back both services with Ceph. With the RBD driver, creating a virtual machine from an image is an instant clone rather than a copy, volumes are replicated across several servers, and losing a disk has no visible impact. Several storage classes can coexist, for example an NVMe pool for databases and a higher-capacity pool for archives.

Keystone: identity and access

Keystone authenticates users and services and organises the cloud into domains, projects and roles. It defines who can create what, in which project and with which quotas. We connect it to your existing directory (LDAP, Active Directory) or identity provider through OpenID Connect or SAML, so your staff sign in with their usual accounts.

Clear governance of projects and roles is essential as soon as several teams or clients share the cloud: we help you define it, with dedicated service accounts for automation and limited, logged administrative access.

Deployment: automated and reproducible

An OpenStack cloud isn't installed by hand. Depending on the context, we deploy it with Kolla-Ansible, which runs each service in containers and uses Ansible for orchestration, or with OpenStack-Helm when the control plane should run on Kubernetes. Either way, all configuration is versioned: the platform can be rebuilt, audited and upgraded predictably.

Deployment follows clear steps: validating the architecture and hardware, setting up the physical network and the Ceph cluster, deploying the highly available control plane, adding compute nodes, load and failover testing, then gradually opening the platform to users with the matching documentation.

Operations: the part that really matters

Most of the work starts after go-live. We monitor every OpenStack service, the message queue, databases, hypervisors and the Ceph cluster with Prometheus and Grafana, and our on-call engineers respond 24 hours a day. Backups of configuration databases and critical data are tested regularly.

OpenStack ships a new release every six months. We plan upgrades with you, validate them first on a staging environment, then apply them service by service to limit impact. We also apply hypervisor and system security patches, manage capacity and add hardware before your projects run short of resources.

  • 24/7/365 monitoring and on-call
  • Planned upgrades tested on staging first
  • Security patches for hypervisors and services
  • Capacity tracking and addition of compute or storage nodes
  • Hardware intervention within 2 to 4 hours on our infrastructure
  • Regular reports and operations reviews

Migrating to OpenStack

Many OpenStack projects start with a migration: from a proprietary virtualisation platform whose licence costs have risen, from a public cloud that has become too expensive, or from ageing physical servers. We inventory existing machines, convert disk images, recreate networks and security rules, then migrate in waves, starting with the least critical environments.

In each wave, applications are tested before the final switchover, with a rollback plan. Your business keeps running throughout the transition.

What LaMeDuSe Cloud takes care of

  • Documented OpenStack architecture (control, compute, network, storage)
  • Automated, reproducible deployment (Kolla-Ansible or OpenStack-Helm)
  • Ceph integration for Cinder, Glance and Nova
  • Neutron networking: private networks, routers, floating IPs, security groups
  • Keystone identity linked to your directory (LDAP, OIDC)
  • Monitoring, backups and planned upgrades
  • Hosting on our hardware or yours
  • Audit and takeover of existing OpenStack clouds

Our methodology

  1. 1

    Context audit

    Volumes, criticality, hardware constraints and goals: context dictates the architecture.

  2. 2

    Tailored architecture

    Argued recommendations, sized for your real loads and peaks.

  3. 3

    Implementation

    Deployment by teams who run the same stack on their own production.

  4. 4

    Handover

    Documentation, training and knowledge transfer for your autonomy.

Technologies

OpenStackKVMCephOpen vSwitch / OVNKolla-AnsibleOpenStack-HelmTerraform

Security

Hardened baseline configuration, secrets managed off the servers, encrypted backups and regular restore tests.

Hosting & infrastructure

Our own production runs on these technologies in our European datacenters: expertise is proven by operations.

Maintenance

Version tracking, security watch and support by the teams who administer the same stack daily.

Why LaMeDuSe Cloud

Our teams work daily with the building blocks an OpenStack cloud rests on: KVM virtualisation, Ceph storage, networking and automation, on our own hardware in our Paris, Frankfurt and Zürich datacenters. It's the same operational rigour we apply for clients such as the French Defence Procurement Agency (DGA) and Free Mobile.

Frequently asked questions

LaMeDuSe Cloud deploys and runs OpenStack clouds and Kubernetes clusters, together or separately, on its European infrastructure or on clients' hardware, with 24/7 monitoring.

Yes, as long as several teams or clients need self-service infrastructure, or you want a public-cloud-style API. For a more compact infrastructure run by a small team, Proxmox VE is sometimes a better fit: we advise based on your context.

Yes. We design the architecture, validate the hardware, deploy the platform in your datacenter and can then operate it remotely. We can also host it on our hardware in Paris, Frankfurt or Zürich.

Yes. We start with an audit (versions, architecture, security, storage health, backups), fix the critical issues, document the platform, then take over its operation and upgrades.

In stages: inventory of machines and dependencies, conversion of disks to the right format, rebuilding networks and security rules in Neutron, then migration in waves with testing and a rollback plan at every step.

We recommend Ceph, which serves Cinder (volumes), Glance (images) and Nova (instance disks), replicates data across several servers and enables live migration of virtual machines.