3.4 Explain the importance of resilience and recovery in security architecture

★ Starred in your notes (Day 5): tabletop exercise — "plan ahead of time" · incremental backup ("like mail coming to pick up your mail") vs. differential ("start Monday, then add other days") · RAID — "don't worry too much on it."

Breakdown 1 — High availability

From class — availability controls: redundancy, fault tolerance, patch management. Availability = server clustering, load balancers, RAID.

Technique	What it is	Difference
Load balancing	Distributes requests across several servers; active/active (all serve) vs. active/passive (standby takes over)	Scales performance and tolerates a node loss
Clustering	Servers act as one system with shared state/storage; failover between nodes	Provides continuity of a service (databases, file servers)
RAID	Disk redundancy — see table below	Survives disk failure, not a backup
QoS (quality of service)	Gives priority to certain traffic (class: "VLAN / VoIP comes up with this")	Keeps voice/critical apps usable under load
Scalability	Scaling up = vertical (bigger box); scaling out/sideways = horizontal (more boxes); elasticity = automatic up and down	Absorbs load (including DDoS)
Redundant power/network	Dual PSUs, dual NICs, dual ISPs	Removes single points of failure

RAID levels (class: "don't worry too much on it" — know the basics):

Level	How	Survives	Note
RAID 0	Striping	Nothing — one disk dies, all data gone	Speed only
RAID 1	Mirroring	One disk	50% capacity
RAID 5	Striping + distributed parity (3+ disks)	One disk	Good balance
RAID 6	Double parity (4+ disks)	Two disks
RAID 10 (1+0)	Stripe across mirrored pairs	One disk per mirror pair	Class diagram: RAID 0 over two RAID 1 pairs (A/A · B/B)

Breakdown 2 — Site considerations

Site	Readiness	Cost	Class note
Hot	Fully equipped, data replicated, minutes to switch	Highest
Warm	Hardware and connectivity in place; restore data/config first	Medium	"warm site → hours to days"; "a compromise"
Cold	Space, power, cooling only; bring everything	Lowest	"Days to weeks"
Mobile	A trailer that can be moved as needed (class)	Varies	Disaster-zone deployments
Geographic dispersion	Sites far enough apart that one disaster can't hit both

Also: platform diversity (use different tech — one exploit shouldn't take everything), multi-cloud (avoid single-provider outage/lock-in), geographic dispersal (class example: AWS regions), redundancy strategies and secure power systems.

From class — Module B continuity planning (the alphabet): BCP business continuity plan (keep the business running) · BIA business impact analysis (what matters, RTO/RPO) · DRP disaster recovery plan (restore IT) · ISCP information system contingency plan (per-system recovery) · COOP continuity of operations plan (government term for keeping essential functions going) · crisis communication plan (who tells whom what) · succession plan (who takes over if leaders are unavailable).

Breakdown 3 — Capacity planning and testing

• Capacity planning: people (staff/skills), technology (licenses, bandwidth), infrastructure (rack space, power) — plan for peak and for disaster load.
• Testing — the class ladder, cheapest to most disruptive: checklist test (walk the plan on paper) → ★ tabletop exercise (talk through a scenario around a table — "plan ahead of time") → simulation test (realistic drill, no production impact) → parallel test (run the backup site alongside production and compare) → full interruption test (actually fail over production — highest risk, highest assurance). CompTIA's list: tabletop, failover, simulation, parallel processing.

Breakdown 4 — Backups

Consideration	Know
Onsite / offsite	Offsite (or cloud) survives site disasters; 3-2-1 rule: 3 copies, 2 media, 1 offsite
Frequency	Driven by RPO — how much data you can afford to lose
Encryption	Backups hold everything; encrypt them (and protect the keys)
Snapshots	Point-in-time image of a VM/volume; fast, but on the same storage
Recovery	Test restores! A backup you haven't restored is a hope, not a control
Replication	Continuous copy to another system/site (near-zero RPO)
Journaling	Log every change so you can roll forward/back to a point in time
Types (class: full · incremental · differential · image · snapshot — "think VM")	Full = everything. Incremental ★ = only what changed since the last backup of any kind — "like the mail carrier picking up just today's mail": fastest backup, slowest restore (full + every incremental). Differential = everything since the last full — "start Monday, then add the other days": grows daily, restore = full + latest differential. Image = whole disk. Snapshot = point-in-time VM/volume state. Class hint: "are they asking for fast backups (incremental) or fast restores (differential)?"

Breakdown 5 — Power

Power events to know (class: "know what these are"): blackout = total loss; brownout = sustained low voltage; spike/surge = brief high voltage; sag = brief low voltage. UPS (battery — bridges seconds to minutes, conditions power) → generator (long outages; takes seconds to start, so you need the UPS too). Also PDUs, dual feeds, managed shutdown.

Breakdown 6 — The metrics (class: "look these up", "study")

Metric	Meaning	Memory hook
RTO — recovery time objective	Max acceptable time to restore service	"How long can we be down?"
RPO — recovery point objective	Max acceptable data loss, measured in time	"How old can the backup be?"
MTTR — mean time to repair/recover	Average time to fix after a failure	Repairable things
MTBF — mean time between failures	Average uptime between failures of a repairable system	Reliability of servers
MTTF — mean time to failure	Expected life of a non-repairable item	"Think of a lightbulb — it dies, you replace it, not repair it" (class)

Exam tip: RPO drives backup frequency; RTO drives site choice (hot/warm/cold). "Replace, don't repair" → MTTF. "Business can tolerate 4 hours of lost data" → RPO = 4h. "Must be back within 1 hour" → RTO = 1h → hot site/replication.

Quick self-check

• Warm site recovery time? (Hours to days.)
• RPO vs. RTO in one line each.
• Load balancing vs. clustering? (Distribution vs. shared-state failover.)
• Why do you still need a UPS if you have a generator? (Generator start-up gap.)