Multi-site architecture
Two separate Digital Realty sites in the Paris area: PAR2 in Aubervilliers and PAR6 in Ivry-sur-Seine.
01 / Virtual machines
Resources matched to your needs.
Size each VM according to your applications, your usage and your performance needs.
The resources of each VM can be adjusted month by month — up or down — as needs evolve: CPU, memory, storage and other associated resources.
Changes made during a month are reflected in the following month’s invoice (M+1).
CPU
Adjustable computing power
RAM
Memory sized to your needs
Storage
Suitable capacity and performance
System
Linux as standard
— Sizing tailored to each project.
Hosting — France · Paris · Digital Realty
02 / Infrastructure
Infrastructure spread across several sites.
Discover where the equipment is located and the principles of physical redundancy.
Physical redundancy
The infrastructure is designed with no single point of failure: redundant power supplies and network links, duplicated hypervisors and network equipment. Data is replicated across three nodes within the Ceph architecture.
Distribution across sites
Roles are separated across the sites: production, backup and disaster recovery, to limit the impact of a major incident or disaster at one site.
Failover between sites
As part of a Disaster Recovery Plan (DRP), virtual machines are duplicated on the standby infrastructure and kept ready to be restarted so that production can resume in the event of a major disaster.
03 / Connectivity
Connectivity at the heart of your infrastructure.
Multi-carrier connectivity, combining peering and several transit providers to avoid any dependence on a single network.
Multiple carriers
Nine IX peering and several transit providers: Telia / Arelion, OpenTransit / Orange and Moji / Cogent.
Network redundancy
Diversified links, carriers and network paths to keep connectivity running.
Independent IP addressing — PI
Pilot’s own IP block, independent of the carriers, announced through several multi-homed BGP routes to preserve flexibility and service continuity.
04 / High availability
Keeping your services running.
Two levels of high availability can be implemented: automatic VM recovery if a hypervisor fails, or real-time failover for the most critical services.
Hypervisor high availability
With Proxmox, a hypervisor failure is detected automatically and the affected VMs can be restarted on another available node of the cluster.
Distributed storage
VM data relies on a redundant Ceph architecture, replicated across several nodes. It remains available when a physical server goes down.
Real-time failover
For critical architectures, services can be replicated across several VMs. Heartbeat and floating IP mechanisms automatically redirect traffic to an available instance in the event of a failure.