Resilience and recovery

Systems fail; good architecture plans for it. This objective covers high availability, site considerations, platform diversity, multi-cloud, continuity of operations, capacity planning, testing, backups, and power.

High availability and sites

High availability keeps services running through component failures: load balancing spreads work across nodes, while clustering lets nodes take over for one another. Recovery sites give somewhere to go when a whole location fails.

Geographic dispersion places sites far enough apart that one disaster doesn't hit both. Platform diversity (different vendors or operating systems) and multi-cloud designs avoid a single flaw or provider outage taking everything down.

Backups

Backups are only useful if they can be restored. Key choices: onsite versus offsite, frequency, encryption, snapshots, replication, and journaling. Replication keeps a copy current, which helps availability but also copies corruption or ransomware instantly; point-in-time backups and snapshots give you something to roll back to.

Continuity, testing, and power

Continuity of operations planning keeps essential functions going during a disruption. Capacity planning covers people, technology, and infrastructure for peak demand. Plans are tested through tabletop exercises (discussion), fail-over tests, simulations, and parallel processing (running the backup system alongside production).

• Uninterruptible power supply (UPS): short-term battery power and clean power during brief outages.
• Generator: longer-term power for extended outages, started while the UPS carries the load.

Key terms

Hot site

A fully equipped recovery site with current data that can take over almost immediately.

Warm site

A recovery site with hardware and connectivity ready, but data must be restored before use.

Cold site

A recovery location with space and power but little or no equipment or data.

Geographic dispersion

Placing sites and resources far enough apart that a single disaster can't affect all of them.

Clustering

Grouping servers so that if one node fails, another takes over its workload.

Platform diversity

Using different vendors, operating systems, or technologies so one flaw can't take down everything.

Multi-cloud

Using more than one cloud provider to avoid dependence on a single provider.

Continuity of operations (COOP)

Planning that keeps essential business functions running during and after a disruption.

Capacity planning

Ensuring enough people, technology, and infrastructure exist to meet demand, including peaks.

Tabletop exercise

A discussion-based walk-through of a scenario to test a plan without touching systems.

Fail-over test

Deliberately switching operations to a backup system to prove it can take over.

Snapshot

A point-in-time copy of a system or volume's state that can be restored later.

Replication

Continuously copying data to another location so a current copy is always available.

Uninterruptible power supply (UPS)

A battery unit that provides immediate short-term power and conditioning during outages.

Generator

An engine-driven power source that supplies electricity during extended outages.