TL;DR
This paper proposes a new reference architecture for blockchain-based traceability systems (BTS) that solves the "complexity vs. performance" bottleneck. By combining Domain-Driven Design (DDD) and Microservices with a multi-chain blockchain structure, the authors created a system that is highly extensible, easier to maintain, and capable of handling high-volume supply chain data without the typical performance lag of a single-chain ledger.
What Problem Does This Paper Address?
Global supply chains are massive and prone to crises, like food safety scandals. While blockchain is a natural fit for traceability because it provides an immutable, transparent record, implementing it at scale is a nightmare.
The authors identify three primary hurdles:
- Complexity: Business processes in supply chains are broad and frequently change. Updating a monolithic blockchain system is difficult and error-prone.
- Data Overload: Storing every piece of business data on a single blockchain leads to performance degradation and massive storage overhead.
- Centralization Risks: Traditional systems often rely on centralized databases that are easy to manipulate and represent a single point of failure.
Methods
The authors propose a layered reference architecture that moves away from the "one-size-fits-all" blockchain approach. Their method relies on three pillars:
- Domain-Driven Design (DDD): They use DDD to break the complex supply chain into "Bounded Contexts." Instead of one giant system, they identified six distinct domains: User Authority, Enterprise Record, Warehousing, Inventory, Supervision, and Traceability.
- Microservices Architecture (MSA): Each domain is implemented as an independent microservice. This allows developers to update the "Warehousing" logic without touching the "User Authority" code, greatly improving system agility.
- Master-Slave Multi-Chain Network: This is the core technical innovation. Rather than one blockchain, they use a Main Chain to store metadata and indexes, and six specialized Sub-Chains (one for each domain). Transactions are processed in parallel across these sub-chains, preventing bottlenecks.
- Hybrid Storage: To keep the blockchain lean, they use IPFS (InterPlanetary File System) for large documents like certificates and a relational database for non-critical data, only storing hashes on-chain.
Datasets
The researchers did not use a standard public dataset. Instead, they collaborated with a blockchain research institute in China to import real-world data from a salmon supply chain scenario. This included tracking imported Norwegian salmon through various stages of the supply chain, involving over 5,000 enterprises.
Key Insights and Ideas
- Sharding by Business Logic: The most significant "aha moment" is applying the concept of sharding not just to data, but to business domains. By giving each domain its own sub-chain, the system achieves natural load balancing.
- The "Anti-Corrosion" Layer: In their DDD implementation, they utilized an anti-corrosion layer in the supervision context. This prevents changes in one domain (like a new warehouse regulation) from "polluting" or breaking the logic in other domains.
- Smart Contract Granularity: The authors argue that smart contracts should match the granularity of domain entities. This ensures that when business requirements change, only a specific, small contract needs to be redeployed, rather than the entire system.
Key Results
The architecture was evaluated using the Architecture Trade-off Analysis Method (ATAM) and a prototype built on Hyperledger Fabric.
- Throughput: The system achieved a querying throughput of approximately 2800 TPS (Transactions Per Second).
- Latency: The average latency was clocked at 1.4 seconds, which is well within the requirements for real-time supply chain monitoring.
- Extensibility: Evaluation showed that business logic changes required less than 2% code modification, proving the system is highly adaptable to new business requirements.
- Security: By using sub-chains, data privacy is improved; only organizations registered to a specific sub-chain can access its ledger.
Limitations and Future Work
While the multi-chain approach solves performance issues, the authors acknowledge a few trade-offs:
- Interoperability: Storing data off-chain (in IPFS) improves speed but creates a "sacrifice in interoperability" compared to keeping all data directly on the chain.
- Complexity of Management: Managing six sub-chains and a main chain is inherently more complex than managing a single ledger.
- Future Directions: The authors plan to explore better interoperability between different reference architectures and aim to scale the prototype to support even more diverse real-world scenarios beyond the salmon industry.
Summary updated