How snu-dbs/m2bench works
High-Level Components
a. Dataset Management Layer
• Data Generation Components
• Data Scaling Components
• Raw Dataset Storage
• Generated Dataset Storage
b. Implementation Layer
• Multiple DBMS Implementations
• Query Execution Components
• Data Loading Components
c. Task Execution Layer
• Benchmark Tasks
• Performance Measurement
• Results Collection
Component Relationships and Data Flow
[Raw Datasets] → [Data Generator] → [Data Scaler] → [Generated Datasets]
• ↓
[Implementation Layer] ← [Data Loading Scripts] ← [Generated Datasets]
• ↓
[Task Execution] → [Performance Results]
Specific Design Elements
• Hexagons for data processing components (Generator, Scaler)
b. Implementation Components:
• Separate containers for each DBMS (AgensGraph, ArangoDB, Polyglot)
• Connection points between different data models
• Data loading workflows
• Query execution paths
c. Task Execution:
• Task categories
• Multi-model interactions
• Performance measurement points
Technical Details
a. Data Model Support:
• Relational (MySQL/PostgreSQL)
• Document (MongoDB)
• Property Graph (Neo4j)
• Array (SciDB)
b. Integration Points:
• Data format conversions
• Cross-model query execution
• Performance monitoring hooks
Implementation-Specific Details
For each DBMS implementation, show:
• Data loading pipelines
• Query execution paths
• Model-specific optimizations
• Cross-model query handling
Additional Elements
Include:
• Version information for supported DBMSs
• Data flow direction indicators
• System requirements and dependencies
• Scale indicators for the benchmark
Remember to include a clear indication of the three main datasets (ecommerce, healthcare, disaster) and how they flow through the system, from raw data to benchmark execution.
This design will help developers understand both the high-level architecture and the detailed interactions between components in this multi-model benchmark system.