메뉴
GCP cost engineering and financial operations. Analyze billing, optimize resources, manage budgets, reduce spending. Use when discussing GCP costs, budget optimization, pricing decisions, cost analysis, or spend management.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | gcp-finops |
| description | GCP cost engineering and financial operations. Analyze billing, optimize resources, manage budgets, reduce spending. Use when discussing GCP costs, budget optimization, pricing decisions, cost analysis, or spend management. |
| allowed-tools | Read, Grep, Glob, WebFetch |
| kb-sources | ["wiki/software-engineering/gcp-finops"] |
| updated | "2026-04-19T00:00:00.000Z" |
Cost engineering and financial operations patterns for Google Cloud Platform.
Note: Pricing estimates are approximate and based on us-central1 region. Use the GCP Pricing Calculator for current rates.
┌─────────────┐
│ INFORM │ ← Visibility into spending
│ (analyze) │
└──────┬──────┘
│
┌──────▼──────┐
│ OPTIMIZE │ ← Right-size, eliminate waste
│ (reduce) │
└──────┬──────┘
│
┌──────▼──────┐
│ OPERATE │ ← Maintain efficiency
│ (control) │
└─────────────┘
| Config | Cost/hour | Monthly (24/7) | Use Case |
|---|---|---|---|
| 0.5 vCPU, 256MB | $0.0144 | $10.50 | Minimal API |
| 1 vCPU, 512MB | $0.0324 | $23.60 | Standard API |
| 2 vCPU, 1GB | $0.0648 | $47.20 | Heavy processing |
Key insight: Scale-to-zero saves 80-95% for variable workloads.
| Config | Cost/GPU-sec | GPU Cost/hour | GPU Monthly (24/7) | Use Case |
|---|---|---|---|---|
| 4 vCPU, 16GB + L4 (min) | $0.0001867 | ~$0.67 | ~$490 | Light GPU inference |
| 8 vCPU, 32GB + L4 (rec) | $0.0001867 | ~$0.67 | ~$490 | Production GPU inference |
Note: Table shows GPU cost only. vCPU/memory charges are additional (see Compute table above). Total cost for 8 vCPU/32GB + L4 is ~$0.67 GPU + ~$0.19 vCPU/mem = ~$0.86/hr.
Key constraints:
Caution: Agent-generated GPU cost estimates are frequently wrong (e.g., confusing $/hr with $/min — a 60x error). Cross-check against Cloud Run pricing before committing to a budget — unverified estimates have shipped with 60x errors that made the FinOps recommendation actively harmful.
| Tier | Monthly | RAM | Best For |
|---|---|---|---|
| db-f1-micro | $16 | 0.6GB | Dev only |
| db-g1-small | $28 | 1.7GB | Staging |
| db-standard-1 | $68 | 3.75GB | Small prod |
| db-standard-2 | $136 | 7.5GB | Medium prod |
Key insight: +35% for HA (REGIONAL).
Pricing model: $0.049/GB/hour (BASIC tier, us-east4)
| Size | BASIC | STANDARD_HA |
|---|---|---|
| 1GB | ~$35/mo | ~$70/mo |
| 2GB | ~$72/mo | ~$144/mo |
| 5GB | ~$180/mo | ~$360/mo |
Key insights:
Cost calculation formula:
Monthly cost = Memory_GB x $0.049/hour x 730 hours/month x HA_multiplier
Where:
HA_multiplier = 1.0 (BASIC) or 2.0 (STANDARD_HA)
Example: 2GB STANDARD_HA
= 2 x $0.049 x 730 x 2.0
= $143.08/month
| Current State | Recommendation | Savings |
|---|---|---|
| CPU <10% utilization | Downsize tier | 30-50% |
| Memory <30% used | Reduce RAM allocation | 20-40% |
| Scale-to-zero disabled | Enable for dev/staging | 80-95% |
| HA in staging | Disable HA | 26-35% |
| SSD in staging | Switch to HDD | 47% |
| Threshold | Action | Audience |
|---|---|---|
| 50% | Informational | Team |
| 75% | Review spending | Tech Lead |
| 90% | Reduce non-essential | Team Lead |
| 100% | Emergency review | Management |
See reference.md for detailed pricing research methods and optimization patterns.
Sprint planning patterns including effort estimation, cross-project validation, ticket refinement workflows, two-validator parallel intent-only refinement, Foundation→Gates ticket pairs, workstream split thresholds, and hub/per-project reconciliation. Use for sprint capacity planning, ticket validation, dependency discovery, tier-gated dispatch, strategic pivots, or coordinating multi-agent ticket processing.
Multimodal AI patterns for extracting structured information from video using Gemini, GPT-4o, and Claude vision. Frame sampling strategies, structured output with Pydantic, temporal reasoning, and cost optimization. Use for video analysis, report generation from video, and visual inspection.
Apple Foundation Models framework for on-device AI. LanguageModelSession, @Generable macro, structured output, token management, and availability checking.
Tactical DDD patterns for clean Python architecture including Domain Model, Repository, Service Layer, Unit of Work, and Aggregates. Use for backend architecture decisions, implementing data access layers, separating business logic from infrastructure, or reviewing code separation.
Architecture Decision Records (ADRs) for documenting significant technical decisions. Use for capturing decision context, rationale, consequences, and maintaining decision history.
Architecture visualization patterns using Mermaid and ASCII diagrams. Use for service flows, sequence diagrams, C4 models, decision matrices, or expressing architectural ideas visually.