Disaster Recovery Planning: From RPO/RTO to the Test Calendar

In most organisations the disaster recovery plan (DRP) sits on a shelf as a written document. The real question is whether the numbers in that document are actually achievable. A reliable DR plan is built in three steps: business impact analysis, setting targets and regular testing.

1. Business impact analysis (BIA)

Calculate the cost of each business process stopping. For a company with annual revenue of 262.8 million TRY that sells mostly online, hourly revenue is about 30,000 TRY. An order system down for 24 hours means 720,000 TRY of direct lost revenue, not counting reputation and customer loss.

2. RPO and RTO targets

• RPO: how much data loss can you accept at most? It may be 5 minutes for the order database and 24 hours for a file server.
• RTO: how soon at the latest must the system be running? It may be 1 hour for the order system and 3 days for an internal reporting system.

The tighter the target, the higher the cost. So instead of tying all systems to one standard, three tiers are more economical: critical (RPO ~ 0), important (RPO 15 minutes - 1 hour) and standard (RPO 24 hours).

3. Match the architecture to the target

• Critical systems: synchronous replication to a second site and automatic failover
• Important systems: replication every few minutes and semi-automatic failover
• Standard systems: daily backup, immutable copy, cloud target

4. An untested plan is not a plan

The DR plan should be tested at least twice a year without affecting production. If the real RTO measured in a test is far from the target RTO, the plan or architecture is corrected. For a ransomware scenario, "restore into a clean environment" must be tried separately.

5. People and communication plan

DR is not only technical. Who decides, who informs customers and employees through which channel, where are suppliers' emergency numbers? This information must be written down and accessible offline.

Cost

In this example company a three-tier DR solution adds about 1.2 million TRY a year in running cost; roughly half the cost of a single 24-hour outage. The figures are examples; we recommend calculating the return on investment with your own BIA data.