Even though we spend quite some time with our head in the 'cloud', BIT is very grounded. We are aware that a reliable infrastructure is essential in creating a stable surrounding, regardless of the label that is attached later: cloud, VSP (virtual private server) or virtual server.
Virtual servers have many advantages, like the sharing of resources, cloning and high accessibility rates due to fail-over techniques. BIT offers 'resource pools' as well as separate servers on its virtualisation platform. A resource pool provides access to a fixed amount of memory, processors and disc space. These resources can be divvied up by the clients and assigned to a number of virtual servers.
Like physical servers, virtual servers can be load balanced. BIT's load balancing enables an application to be supported by multiple servers. Virtual servers behind our load balancers are activated in two geographically separated data centers. If one of the servers fails, either a physical or a virtual one, the application will remain available.
All virtual servers on BIT's infrastructures are supplied as high availability servers. In case a physical server supporting multiple virtual servers fails, the affected virtual servers are transferred to another physical server. BIT always strives for high availability after all.
That high availability manifests itself in the arrangement of storage for virtual servers as well: no local drives on the physical server, but central storage. And not in a single central storage, but with three Ceph replicas in our storage cluster. All virtual servers are - of course - always on three storage systems in three different data centers.