DBT – Digital Box Technologies ·
At compute level, assess a cluster architecture suited to availability objectives, with sufficient spare capacity and validated quorum behaviour. Virtual machine restart depends on configuration and remaining resources.
For storage and networking, double the paths: redundant controllers, aggregated links, stacked or paired switches. A single cable should never be able to stop production.
Power backup, cooling and room monitoring are often overlooked, yet they cause many incidents.
Finally, test. A resilient architecture is one whose behaviour has been checked by simulating failures, and whose procedures are written down.
Start with a service, not a hardware list
Map a critical application's path through user devices, networking, firewall, directory, hypervisor, storage and power. Ask what happens when each dependency stops. Two servers connected to the same failed component are not two independent service paths. This map directs investment toward failures that would actually interrupt the business.
Plan for degraded operation
Capacity is not only about normal operation. Check that remaining resources can support priority workloads during failure or maintenance. Cluster behaviour, quorum and storage must be assessed together. Node count alone does not guarantee high availability: configuration, spare capacity and testing still matter.
Test without improvising an incident
An exercise needs scope, authorisation, named participants and a way to stop safely. Begin with planned maintenance, then assess a link or component failure under controlled conditions. Record alerts, recovery time, manual actions and user-visible impact. A green dashboard is not proof that business operations continued.
Keep a separate recovery strategy
Redundancy improves availability for certain failures; it does not replace backups or disaster recovery. Corruption or mistakes may propagate across redundant components. Retain protected copies and restoration procedures. For a site in Rabat or elsewhere in Morocco, document actual power, intersite connectivity and physical access constraints rather than assuming identical conditions everywhere.