3.3 - Disaster Recovery

Disaster recovery plan (DRP)

• Detailed plan for resuming operations after a disaster

– Application, data center, building, campus, region, etc.

• Extensive planning prior to the disaster

– Backups

– Off-site data replication

– Cloud alternatives

– Remote site

• Many third-party options

– Physical locations

– Recovery services

Recovery

• Recovery time objective (RTO)

– Get up and running quickly

– Get back to a particular service level

• Recovery point objective (RPO)

– How much data loss is acceptable?

– Bring the system back online; how far back

does data go?

• Mean time to repair (MTTR)

– Time required to fix the issue

• Mean time between failures (MTBF)

– Predict the time between outages

Site resiliency

• Recovery site is prepped

– Data is synchronized

• A disaster is called

– Business processes failover to the alternate

processing site

• Problem is addressed

– This can take hours, weeks, or longer

• Revert back to the primary location

– The process must be documented for both directions

Cold site

• No hardware - Empty building

• No data - Bring it with you

• No people - Bus in your team

Hot site

• An exact replica - Duplicate everything

• Stocked with hardware

– Constantly updated - You buy two of everything

• Applications and software are constantly updated

– Automated replication

• Flip a switch and everything moves

– This may be quite a few switches

Warm site

• Somewhere between cold and hot

– Just enough to get going

• Big room with rack space

– You bring the hardware

• Hardware is ready and waiting

– You bring the software and data

Tabletop exercises

• Performing a full-scale disaster drill can be costly

– And time consuming

• Many of the logistics can be determined through analysis

– You don’t physically have to go through a

disaster or drill

• Get key players together for a tabletop exercise

– Talk through a simulated disaster

Validation tests

• Test yourselves before an actual event

– Scheduled update sessions (annual, semi-annual, etc.)

• Use well-defined rules of engagement

– Do not touch the production systems

• Very specific scenario - Limited time to run the event

• Evaluate response - Document and discuss