BRIEF / 02.A.01
Where the platform can stop working
A platform can run on several servers and still have a single point of interruption: a database, administrative identity, storage, network or provider. Design starts with the value of the service and the consequences of losing it.
BRIEF / 02.A.02
How much downtime and data loss are acceptable
RTO — Recovery Time Objective — expresses the required recovery time. RPO — Recovery Point Objective — defines the tolerable loss of recent data. Objectives concern the application functions that must return, rather than server startup alone.
BRIEF / 02.A.03
Redundancy, replication and shared failures
HA — High Availability —, load balancing and replication are assessed against actual failure domains. The design addresses database consistency, shared dependencies and network partitions. Replication does not replace recovery from a propagated logical error.
BRIEF / 02.A.04
Configure the platform and recovery procedures
Cerberox configures and integrates load balancing, database HA — High Availability —, replication and backup copies. It prepares failover and recovery procedures and documents the dependencies needed to run them. Servers, storage and networking can come from third-party providers; the documentation specifies access and responsibilities.
| Requirement | Restore agreed application functions against RTO — Recovery Time Objective — and RPO — Recovery Point Objective. |
|---|---|
| Engineering and implementation | Configure redundancy, load balancing, replication, copies and procedures for the failures to withstand. |
| Planned test | Run authorised failure and restore tests; measure restored functions, recovery time and recovered data point. |
| Handover materials | Configurations and dependencies, runbooks, test conditions, results and measured limits. |
BRIEF / 02.A.05
Measure recovery of application functions
Agreed tests observe failure behaviour and the return of application functions. Measured timings and the recovered data point are compared with RTO and RPO. Restore is checked through the usability of the service.
BRIEF / 02.A.06
Management and operating commitments
SLA — Service Level Agreement —, response times, monitoring and escalation are defined where included in the engagement. A test result describes the conditions observed; it does not by itself guarantee behaviour under every possible failure.