Reliability
RoboDesk carries mission-critical customer operations for multi-branch enterprises across the Middle East and Africa. This is how we keep it up.
Service commitments
| Uptime target | 99.9%+ on production instances. |
|---|---|
| Scale | Billions of interactions handled annually across all tenants. |
| Architecture | Kubernetes with horizontal pod autoscaling; each tenant isolated in its own namespace. |
| Deployment model | GitOps. Every change is a reviewed commit, deployed by ArgoCD and revertible. |
| Data stores | MongoDB for operational data, Elasticsearch for long-term history and analytics. |
| Monitoring | Prometheus metrics and Kibana dashboards, with end-to-end synthetic checks across tenants. |
How we build for resilience
Reliability is an engineering practice, not a number on a page. These are the things that actually keep the platform available.
- Tenant isolation. Each customer runs in a dedicated namespace. A noisy tenant cannot degrade another.
- Autoscaling. Pods scale horizontally with load, so campaign spikes and seasonal peaks do not queue.
- Channel failure containment. A provider outage on one channel does not stop the others; interactions queue and resume.
- Progressive delivery. Releases roll out gradually and roll back automatically on failed health checks.
- Continuous validation. Automated end-to-end tests exercise real flows across tenants, not just unit tests in CI.
Support and escalation
Enterprise agreements include a named account manager and a defined escalation path with response targets by severity. Incidents affecting a production tenant are escalated to engineering directly rather than queued behind general support.
Specific availability commitments, response targets and service credits are set in your service level agreement. Ask your account manager, or contact us for a copy.
Where your data lives
You choose the region. RoboDesk runs in Egypt, Saudi Arabia, the GCC and Europe, on Google Cloud, Microsoft Azure, AWS or Huawei Cloud — or entirely on your own infrastructure. See Security for the full picture.
See it run on your own traffic.
Thirty minutes with a CX engineer. No slideware.