Continuity Strategies, DR Architecture & Data Resilience, 10 min read

Ready reckoner

ISO 22301 clause 8.3

Sub-clause Requirement
8.3.2 Identify Options before, during, after disruption that: meet timeframes + capacity · protect · reduce likelihood · shorten duration · limit impact · secure resources
8.3.3 Select Fit to requirements · risk appetite · costs and benefits
8.3.4 Resources People · information/data · buildings, workplaces, utilities · facilities, equipment, consumables · ICT · transport/logistics · finance · partners/suppliers
8.3.5 Implement Solutions maintained and activatable when needed

Option families: diversify · replicate · standby · acquire after event · work around · insure (never alone) · accept (approved).

Selection steps: filter by RTO/RPO/MBCO → test vs scenarios → cost-benefit → risk appetite (fund or board-accept) → record rationale.

People and premises

Strategy Fits Watch
Remote work Cloud-based knowledge work Identity, connectivity, regional power/network loss
Split operations City-level events Cost, skills parity
Work-area recovery Dealing, cash ops, secure rooms Oversubscribed shared seats
Displacement Multi-site firms Rehearse it
Cross-skilling Key-person, mass absence Skills fade

Key-person: deputies · documentation · delegations · break-glass with dual control. Life safety and welfare first.

DR patterns

Pattern RTO RPO Cost
Backup & restore hours–days hours ●
Pilot light hours minutes ●●
Warm standby ≤ 1–2 h minutes ●●●
Hot standby minutes seconds ●●●●
Active-active ≈ 0 ≈ 0 ●●●●●

Replication copies corruption → every pattern needs isolated immutable backups.

Sync = zero loss, short distance (latency). Async = distance, some loss. Three-site: near site (sync) + far DR (async). SEBI MIIs: near site near-zero RPO; DR staffed, runs independently; live from DR twice yearly.

Cloud layers

Layer Protects Does not protect
Multi-AZ Hardware, data hall, power Region, control plane, ransomware
Multi-region Regional event Provider-wide, shared IAM, ransomware
Multi-cloud Provider failure, exit Ransomware in shared tooling
Isolated immutable copies Ransomware, admin compromise, corruption — (needs tested restore)

Shared responsibility: provider = resilience of the cloud; you = resilience in the cloud. SLA credits ≠ recovery.

Hidden dependencies: identity/SSO · DNS · keys/HSM · certificates · secrets · monitoring · CI/CD · partner IP whitelists.

Failover decision: criteria · named decider + deputies · time limit (SEBI MII: declare in 30 min) · cyber check (is DR clean?). Fail-back = planned change.

Data resilience

3-2-1-1-0: 3 copies · 2 media · 1 off-site · 1 immutable/air-gapped · 0 restore errors.

Survive an attacker: separate account/domain · separate creds + hardware MFA · object lock (compliance mode) · multi-person approval for retention · isolation · monitoring (deletion, size anomaly) · protected keys · retention depth beyond likely dwell time.

Clean room (IRE): vault → choose known-good point → restore isolated → scan + validate integrity → rebuild + rotate secrets → promote in stages.

Integrity: reconcile with independent source of truth · control totals · mass-change detection.

Test Typical Tier 1 frequency
Job monitoring Daily
File/record restore Monthly
Full restore, timed vs RTO Quarterly
Clean-room exercise Yearly
Immutability test Yearly

Suppliers and exit

Options: require supplier resilience · multi-source · warm alternative · in-house fallback · customer-side workaround · escrow · planned exit.

Contract terms: RTO/RPO · BCP/DR testing + results · joint tests · incident notice (e.g. 2 h) · audit + regulator access · sub-contracting · data location/return · exit assistance · termination rights.

Exit plan: triggers · immediate continuity (MBCO) · routine data copies · transition options + timelines · roles · communications · tested. Stressed exit (sudden) vs planned exit (months, parallel run). "Re-tender" ≠ exit plan.