2nd/10_Wiki/Topics/AI_and_ML/Enterprise-Software-Engineering.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-enterprise-software-engineering	Enterprise Software Engineering	10_Wiki/Topics	verified	self	enterprise SE SDLC large-scale software enterprise architecture	none	A	0.95	applied	software-engineering enterprise sdlc process architecture scale		2026-05-10	pending	language applicable_to Universal Enterprise Large-scale Regulated

Enterprise Software Engineering

매 한 줄

"매 scale + 매 compliance + 매 long-lifecycle 의 software". 매 startup velocity 의 X — 매 audit + 매 SLA + 매 multi-team. 매 modern: 매 platform engineering + DevSecOps + observability + AI-augmented (Copilot, Cursor, Anthropic).

매 핵심

매 challenge

• Scale: 매 100s of teams.
• Compliance: SOX, HIPAA, GDPR, PCI.
• Legacy: 매 monolith + 매 mainframe.
• Multi-stakeholder: 매 product + ops + security + legal.
• Long lifecycle: 매 10년+.

매 modern paradigm

• Platform engineering: 매 IDP (Internal Dev Platform).
• DevSecOps: 매 security shift-left.
• GitOps: 매 declarative infra.
• SRE: 매 reliability budget.
• DORA metrics: 매 4 key.
• Team Topologies: 매 stream-aligned + platform.

매 architecture

• Monolith → Microservice: 매 strangler.
• Modular monolith: 매 alternative.
• Event-driven: 매 Kafka, EDA.
• API-first: 매 OpenAPI, gRPC.
• Data mesh: 매 domain-owned data.

매 process

• Agile / Scrum: 매 small batch.
• SAFe: 매 enterprise scale (controversial).
• Trunk-based: 매 modern CI/CD.
• DORA: 매 deploy frequency, lead time, MTTR, change fail rate.

매 응용

1. Banking: 매 core system.
2. Telco: 매 BSS / OSS.
3. Healthcare: 매 EHR.
4. Government: 매 procurement.
5. Insurance: 매 claim.

💻 패턴

DORA metrics dashboard

def compute_dora(deployments, incidents, period_days=30):
    return {
        'deploy_frequency': len(deployments) / period_days,
        'lead_time_p50_hours': median(d.commit_to_deploy_hours for d in deployments),
        'mttr_minutes': median(i.detect_to_resolve_min for i in incidents),
        'change_fail_rate': sum(d.caused_incident for d in deployments) / len(deployments),
    }

Strangler fig (legacy migration)

// 매 facade routes new 의 new, old 의 old
async function getUser(id: string) {
  if (await featureFlag('new-user-service', { userId: id })) {
    return newUserService.fetch(id);
  }
  return legacyUserDao.findById(id);
}

Platform IDP (Backstage)

# catalog-info.yaml
apiVersion: backstage.io/v1alpha1
kind: Component
metadata:
  name: payments-service
  annotations:
    backstage.io/source-location: url:https://github.com/acme/payments
spec:
  type: service
  lifecycle: production
  owner: team-payments
  system: checkout

GitOps (ArgoCD)

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: payments-prod
spec:
  source:
    repoURL: https://github.com/acme/k8s-manifests
    path: prod/payments
  destination:
    server: https://kubernetes.default.svc
    namespace: payments
  syncPolicy:
    automated: { prune: true, selfHeal: true }

SLO / Error Budget

# 매 99.9% SLO → 43.2 min/month error budget
slo:
  service: payments
  objective: 99.9%
  measurement_window: 30d
  burn_rate_alerts:
    - severity: page
      lookback: 1h
      threshold: 14.4  # 매 fast burn (consume 1d in 1h)
    - severity: ticket
      lookback: 6h
      threshold: 6

Audit log

function audit(action: string, actor: string, target: string, metadata: object) {
  auditStream.publish({
    timestamp: new Date().toISOString(),
    action, actor, target,
    metadata,
    correlationId: getRequestId(),
  });
}
// 매 immutable + retention 7y

Compliance check (PII access)

def access_pii(user_id, requester):
    if not has_role(requester, 'pii_reader'):
        raise PermissionError()
    audit('pii_read', requester, user_id, {})
    if requires_purpose(user_id):
        return prompt_for_purpose(requester)
    return fetch_user(user_id)

Multi-tenancy (Postgres RLS)

ALTER TABLE orders ENABLE ROW LEVEL SECURITY;
CREATE POLICY tenant_isolation ON orders
  USING (tenant_id = current_setting('app.tenant_id')::uuid);

Trunk-based deploy

on: { push: { branches: [main] } }
jobs:
  deploy:
    steps:
      - run: npm test
      - run: npm run build
      - run: deploy.sh staging
      - run: smoke-test.sh staging
      - run: deploy.sh canary 5
      - run: monitor.sh canary 5m
      - run: deploy.sh prod

Disaster recovery test

def chaos_dr_test():
    # 매 quarterly DR drill
    primary_db.simulate_failure()
    assert app_reads_from(replica_db)
    promote(replica_db)
    assert app_writes_to(replica_db)
    rollback()

Architecture decision record

# ADR-0042: Adopt Kafka over RabbitMQ for event bus

## Context
50 services, growing 5/quarter, current RabbitMQ at 80% capacity.

## Decision
Kafka MSK with mTLS, schema registry, 7-day retention.

## Consequences
+ Replay capability
+ Throughput headroom
- Operational complexity
- Cost +30% Year 1

Team Topologies (boundary)

teams:
  - name: payments
    type: stream-aligned
    owns: [payments-service, billing-svc]
  - name: platform
    type: platform
    provides: [k8s, observability, secrets]
    serves: [payments, checkout, ...]
  - name: security
    type: enabling
    enables: [...]

AI-augmented dev (Copilot policies)

ai_policy:
  copilot: enabled
  data_residency: eu-west-1
  excluded_paths:
    - secrets/
    - compliance/
  audit_log: true

매 결정 기준

상황	Approach
Greenfield	Cloud-native + IDP
Legacy modernize	Strangler fig
Compliance-heavy	DevSecOps + audit
Multi-team	Platform engineering
Reliability	SLO + error budget
Slow deploys	Trunk-based + CI/CD

기본값: 매 platform IDP + 매 GitOps + 매 SLO + 매 DORA tracking + 매 trunk-based + 매 ADR.

🔗 Graph

• 변형: Platform-Engineering · CI/CD Pipeline & IDE Security Integration · SRE
• 응용: Microservices · Modular Monolith · GitOps
• Adjacent: Team Topologies · DORA · Backstage · ArgoCD · Development Communication Standards

🤖 LLM 활용

언제: 매 large org. 매 regulated industry. 매 long-lived system. 언제 X: 매 startup MVP. 매 throwaway.

❌ 안티패턴

• Process for process sake: 매 velocity ↓.
• Skip compliance: 매 fines.
• Big-bang migration: 매 risk.
• Single team owns all: 매 bottleneck.
• No DORA measurement: 매 improvement 의 invisible.

🧪 검증 / 중복

• Verified (Accelerate, Team Topologies, Google SRE, Platform Engineering).
• 신뢰도 A.

🕓 Changelog

날짜	변경
2026-04-20	Auto-reinforced
2026-05-08	Phase 1
2026-05-10	Manual cleanup — DORA + 매 strangler / Backstage / GitOps / SLO / RLS code