Most companies have invested a lot of time, energy and therefore money in setting up an industrialised global backup solution, often coupled with job schedulers to manage infrastructure backup.
Discussions with teams in charge of this area resulted in several stunning discoveries: virtualisation did not spur a re-examination of the solution, the associated strategy, or even the philosophy that led the company toward the decision it made in terms of a backup solution.
The result? Backed-up virtualised platforms with physical platform solutions, an agent policy on each virtual machine (VM), etc. Everything has been carefully transmitted on tape through an infrastructure's intermediary that is as poorly suited to the task as it is costly.
Platform virtualisation involves re-thinking all related services (e.g., data backup, system recovery, monitoring, and rollout). The widespread adoption of D2D2T (see part 1) is a clear choice, thanks to the solution's flexibility and high level of performance. The development of virtual systems resulted in new solutions to address the services mentioned above. These solutions are radically different from their physical environment predecessors. Aside from native data deduplication, the most obvious distinction is the treatment of VMs for exactly what they are--a flat group of files. The idea of backing up an operating system is abandoned in favour of backing up a flat group of files. Files are checked for consistency in advance using snapshot technology, which is common in the virtual environment. This method enables recovery of a VM--in whole or part (with a different granularity for each editor)--and sometimes of a file, even within the guest OS.
The speed and flexibility of D2D recovery cannot be compared with tape management, even though externalisation is still covered by media of this type (the replication of inter-site SAN bays may, however, change the situation rather quickly for everything except the archive portion).
Finally, D2D lets the virtual platform's production teams gain independence and, therefore, become more responsive for recent recoveries, without necessarily affecting the traditional tape backup perimeter, which is mostly managed by a special team.
There is therefore a real need to re-think data backup with virtualised infrastructure. In part 3, we'll see how best to treat VMs according to their criticality.
Ingénieur Avant-Ventes chez NEOCLES, ma mission consiste en du conseil en avant-ventes et études d'opportunités de projets de transformation d'infrastructures et/ou poste de travail.