Graph Database Configuration

Cosmos DB Gremlin / Neo4j Aura Gov #

---

1 Purpose #

The EA 2.0 graph database is not just a data store — it’s the semantic core of the enterprise.
It stores every node, relationship, event, and confidence score so the reasoning engine can understand how the enterprise operates, why outcomes occur, and where to intervene.

---

2 Platform Options #

Option	Type	Ideal For	Notes / Integration
Azure Cosmos DB (Gremlin API)	Managed PaaS	Government / Sovereign deployments	Fully integrated with Azure RBAC + VNet; schema-optional JSON.
Neo4j Aura Gov / Enterprise	SaaS or Self-Managed	Private sector or federated tenants	Advanced Cypher support, graph algorithms, Bloom visualization.

Both are interchangeable under EA 2.0’s data-access layer; the model is portable.

---

3 Schema Philosophy #

EA 2.0 uses a “semistructured ontology”:

• Fixed core labels (Capability, Application, Data, Risk, Control, Outcome).
• Flexible extensions via metadata properties.
• Relationship types carry direction, weight, and evidence.

This design allows schema-on-read flexibility while keeping query performance predictable.

---

4 Core Label and Index Setup (Neo4j Example) #

CREATE CONSTRAINT capability_id IF NOT EXISTS
FOR (c:Capability) REQUIRE c.id IS UNIQUE;

CREATE CONSTRAINT app_id IF NOT EXISTS
FOR (a:Application) REQUIRE a.id IS UNIQUE;

CREATE INDEX FOR (r:Risk) ON (r.severity);
CREATE INDEX FOR ()-[rel:SUPPORTED_BY]-() ON (rel.confidence);

Tip: Use numeric IDs + ULIDs to keep joins efficient.

---

5 Graph Loading Pattern #

1. Stage incoming JSON to Blob Storage.
2. Transform with Function app → normalized YAML/JSON graph payload.
3. Load via Cypher / Gremlin bulk upsert.

Cypher Example

UNWIND $records AS r
MERGE (a:Application {id:r.app_id})
ON CREATE SET a.name=r.app_name, a.created_at=timestamp()
MERGE (c:Capability {id:r.cap_id})
MERGE (a)-[:SUPPORTED_BY {confidence:r.conf}]->(c);

---

6 Performance Tuning Guidelines #

Category	Practice	Impact
Partitioning	Split by domain or region	Parallel queries & smaller index footprint
Caching	Use server-side Gremlin or Neo4j page cache	2–5× faster traversals
Query Hints	RETURN LIMIT clauses for UI requests	Prevents large payloads
Batch Ingest	≤ 500 nodes / transaction	Stable RUs in Cosmos
Compression	GZIP payloads from Functions	Lower network latency

---

7 Governance and Security #

Area	Control	Implementation
Access	Azure AD / Entra ID RBAC	Role → Graph scope mapping
Encryption	In-transit TLS 1.2+, at-rest AES-256	Default in Cosmos / Aura
Audit Logs	Query logs + change capture to Blob	Immutable history
Least Privilege	Separate read/write keys per service	Rotation via Key Vault
Data Residency	Region-locked instances	Meets GDPR / UAE Gov SLA

---

8 Graph Algorithms Enabled #

EA 2.0 activates built-in graph analytics to power insight queries:

Algorithm	Purpose	Example Use
Centrality	Find critical applications	“Which apps support most capabilities?”
Community Detection	Identify related functions / teams	Cluster analysis of domains
Shortest Path	Trace risk impact chains	“How does Risk-R17 reach Finance?”
Similarity	Recommend reusable controls	Compliance reuse patterns

---

9 NLQ and Reasoning Integration #

• The Reasoning API translates natural-language questions into Cypher/Gremlin.
• Safe query templates are stored in Prompt Library.
• Results return as JSON for Power BI or React UI rendering.

Example Query:

“Show all capabilities with AI-related risks and their mitigations.”

→ translates to a multi-hop graph query (2 ms average response time on 100k nodes).

---

10 Backups & Disaster Recovery #

DB Type	Native Option	Frequency	Retention
Cosmos DB	Continuous Backup + Point-in-Time Restore	4 hours RPO	30 days
Neo4j Aura	Daily auto-snapshot	24 hours RPO	7–30 days
Self-Managed Neo4j	Cron export via neo4j-admin dump	6 hours	Configurable

---

11 Monitoring Metrics #

Track these KPIs in Azure Monitor / Neo4j Bloom:

Metric	Target	Interpretation
Query Latency (ms)	< 200 avg	Healthy response
CPU / RU Utilization	< 70 %	Stable load
Failed Writes	0	Data integrity
Graph Size	≤ 10 GB per partition	Manageable index
Node Growth Rate	< 5 %/day	Predictable scaling

---

12 Multi-Environment Strategy #

• Dev/Test/Prod graphs with promotion pipeline (Azure DevOps).
• Snapshot export/import for data migration.
• Feature-flag schema changes via metadata version.

---

13 Advantages of Graph Architecture #

✅ Native model of enterprise interdependencies.
✅ Real-time impact analysis (queries in ms).
✅ Foundation for RAG and predictive governance.
✅ Schema-light yet semantically rich.
✅ Easily visualized and narrative-ready.

---

14 Sample Governance Policy in Graph #

MERGE (p:Policy {id:'POL-123', name:'Tagging Enforcement'})
MERGE (r:Risk {id:'RISK-22'})
MERGE (p)-[:MITIGATES]->(r)
MERGE (p)-[:TRIGGERED_BY]->(:Event {type:'NonCompliantResource'})

This link ensures every policy has a traceable trigger and target risk.

---

15 Takeaway #

The graph is the brainstem of EA 2.0 — where insight becomes structure.
When deployed on Cosmos DB or Neo4j Aura Gov, it delivers speed, security, and semantic clarity at enterprise scale.