Files
oci-deal-accelerator/kb/patterns/service-tiering.yaml
2026-03-18 18:03:44 -03:00

107 lines
4.2 KiB
YAML

---
last_verified: 2026-03-16
pattern:
id: service-tiering
name: "Database Service Tiering Model"
category: operations
---
description: >
Maps customer workloads to standardized service tiers based on business
criticality, availability requirements, and compliance needs. Each tier
defines the HA/DR topology, support model, backup strategy, and operational
responsibilities. Derived from Oracle's ECAL transformational framework.
tiers:
platinum:
label: "Platinum"
description: "Mission-critical, zero tolerance for downtime"
example_workloads: ["Core banking", "Payment processing", "Trading systems", "Telco network"]
sla:
uptime: "99.999%"
rto: "0 (near-zero with TAC + RAC)"
rpo: "0 (synchronous replication)"
planned_downtime: "0 (even for upgrades — edition-based redefinition)"
support: "24x7, 10 min response"
architecture:
ha: "RAC (multi-node) + TAC/AC"
dr: "Synchronous ADG + GoldenGate active-active (optional)"
backup: "ZDLRA or Autonomous Recovery Service, continuous"
isolation: "Full physical isolation — dedicated CDB, dedicated infrastructure"
encryption: "TDE + DB Vault + Key Vault + SQL*Net encryption"
oci_mapping:
primary: "ExaCS or ADB-D (for RAC + full control)"
alternative: "ADB-S only if application is TAC-compatible and RPO=0 is met by local ADG"
cost_impact: "Highest — dedicated infra, dual-site, 24x7 support"
gold:
label: "Gold"
description: "Business-critical, tolerable brief outage"
example_workloads: ["CRM", "ERP", "eCommerce", "Risk analytics"]
sla:
uptime: "99.99%"
rto: "1-4 hours"
rpo: "5-15 minutes"
planned_downtime: "4x 4hr/year maintenance windows"
support: "24x7, 4 hour response"
architecture:
ha: "TAC + local ADG (auto-failover)"
dr: "Asynchronous cross-region ADG"
backup: "Autonomous Recovery Service, incremental forever"
isolation: "Dedicated CDB (dedicated home), instance caged (not oversubscribed)"
encryption: "TDE + Key Vault"
oci_mapping:
primary: "ADB-S Serverless (with TAC + ADG)"
alternative: "ExaCS if RAC explicitly needed or >128 OCPU sustained"
cost_impact: "High — primary + standby OCPU cost, cross-region DR"
silver:
label: "Silver"
description: "Important but recoverable from backup"
example_workloads: ["HR systems", "Financials", "Internal portals", "Reporting"]
sla:
uptime: "99.5%"
rto: "1 day"
rpo: "1 day"
planned_downtime: "4x 4hr/year maintenance windows"
support: "24x7, 8 hour response"
architecture:
ha: "Single instance or RAC optional"
dr: "Offsite backup only — no real-time replication"
backup: "Automated daily backup, 30-day retention"
isolation: "Shared CDB (shared binaries) acceptable"
encryption: "TDE recommended, DB Vault optional"
oci_mapping:
primary: "ADB-S Serverless (no ADG needed)"
alternative: "Base DB Service for legacy compatibility"
cost_impact: "Medium — no standby cost, shared infrastructure"
bronze:
label: "Bronze"
description: "Non-critical, best-effort recovery"
example_workloads: ["Dev/Test", "Room booking", "Internal tools", "Sandbox"]
sla:
uptime: "98%"
rto: "Best effort"
rpo: "Best effort"
planned_downtime: "Flexible"
support: "Business hours, 2 day response"
architecture:
ha: "None — single instance"
dr: "Database export or backup to Object Storage"
backup: "Weekly backup, 7-day retention"
isolation: "Shared CDB, oversubscribed OK"
encryption: "TDE optional"
oci_mapping:
primary: "ADB-S Serverless (minimum OCPU, auto-scaling off)"
alternative: "VM DB Standard Edition for cost"
cost_impact: "Lowest — minimal resources, no redundancy"
auto_assignment_rules:
- "If customer says '99.99%' or 'zero downtime' → Platinum or Gold (ask which)"
- "If customer says 'PCI' or 'banking regulations' → minimum Gold"
- "If workload type is 'dev/test' → Bronze unless explicitly stated otherwise"
- "If customer says 'reporting' or 'read-only' → Silver unless SLA says otherwise"
- "If no SLA stated → assign Silver and flag for confirmation"