Graph Database

Graph Database findings

Database Options

Database	Key Features	Pros	Cons
Neo4j	Optimised for relationship-heavy queries and traversals. Uses Cypher query language, which is relatively simple.	Excellent for multi-hop traversals. Strong visualisation tools. Relationship properties natively supported.	Limitations on community edition but for contract tracing etc its probably enough Costs for enterprise edition. Requires setup and hosting.
AWS Neptune	Fully managed service supporting Gremlin and SPARQL.	AWS integration. Scalable and highly available.	Higher cost. Steeper learning curve for SPARQL/Gremlin.
TigerGraph	Distributed graph database for large-scale analytics.	Excellent performance for deep relationships. Handles large datasets well.	Enterprise-focused pricing. Requires learning GSQL.
PostgreSQL + Apache AGE	Extends PostgreSQL to support graph functionality.	Cost-effective. Integrates with existing relational workflows.	Limited advanced graph features. Not natively supported on RDS Requires self-hosting PostgreSQL.
ArangoDB	Combines graph, document, and key-value models.	Flexible for mixed workloads. Cost-effective.	Less specialised for graph-specific tasks.

Neo4J

Neo4J would be suitable for what we're trying to achieve. The community edition is most likely suffice for our dataset and things like relationships,contact tracing etc.

For a Small Dataset (2GB)

1. EC2 with EBS
   • Best For: Simple and low-cost setup.
   • Estimated Cost: ~$9/month.
2. ECS with EFS
   • Best For: Flexibility (though it might be overkill unless scaling significantly).
   • Estimated Cost: ~$16/month.
3. Neo4j Aura Free Tier
   • Best For: Simplicity and no upfront cost (if workload fits within free limits).

For a 5GB Dataset

1. EC2 with EBS
   • Best For: Simple, cost-efficient setup with moderate scaling needs.
   • Configuration:
     • Instance: t3.medium (2 vCPUs, 4GB RAM) recommended for moderate query loads.
     • Storage: 20GB EBS gp3 (to allow room for growth and logs).
   • Estimated Cost: ~$40/month (EC2: ~$38, EBS: ~$2).
2. ECS with EFS
   • Best For: Flexibility, especially if you’re familiar with containerized environments.
   • Configuration:
     • 1 ECS Fargate task (1 vCPU, 2GB RAM) with 10GB EFS.
   • Estimated Cost: ~$25/month (ECS: ~$22, EFS: ~$3).
3. Neo4j Aura Professional
   • Best For: Workloads exceeding free tier limits (200K nodes/400K relationships).
   • Estimated Cost: ~$65/month.

For a 10GB Dataset

1. EC2 with EBS
   • Best For: Cost efficiency with predictable workloads.
   • Configuration:
     • Instance: t3.medium (2 vCPUs, 4GB RAM) or m5.large (2 vCPUs, 8GB RAM) for heavy workloads.
     • Storage: 30GB EBS gp3 for dataset and transaction logs.
   • Estimated Cost: ~$41/month (EC2: ~$38, EBS: ~$3).
2. ECS with EFS
   • Best For: Scalability and AWS integration.
   • Configuration:
     • 1 ECS Fargate task (2 vCPUs, 4GB RAM) with 20GB EFS.
   • Estimated Cost: ~$35/month (ECS: ~$30, EFS: ~$5).
3. Neo4j Aura Professional
   • Best For: Handling datasets larger than the free tier but requires a subscription.
   • Estimated Cost: ~$65/month.