SASLedger: A Secured, Accelerated Scalable Storage Solution for Distributed Ledger Systems

Haoli Sun, Bingfeng Pi, Jun Sun, Takeshi Miyamae, Masanobu MorinagaFuture 2021 · DOI

TL;DR

This paper introduces SASLedger, a novel off-chain storage solution for distributed ledger systems like Hyperledger Fabric. By offloading block data to a remote database while retaining concise commitments on nodes, SASLedger dramatically reduces storage requirements for individual nodes by 93.3% without compromising data integrity. This approach also improves transaction processing speed, boosting throughput by 9.6% compared to a prior solution.

What Problem Does This Paper Address?

Blockchain technology's core strength lies in its distributed, tamper-proof ledger. However, this comes at a cost: every node in the network typically stores a complete copy of all transactions and blocks. As blockchains grow, this leads to an ever-increasing storage burden, potentially hindering decentralization by making it difficult for individuals to run nodes. This paper tackles this "storage scalability issue," particularly for permissioned blockchains like Hyperledger Fabric, aiming to reduce the storage footprint on individual nodes while maintaining the integrity and accessibility of the data.

Methods

SASLedger employs a two-phase approach to manage storage and performance:

  • Transaction-Processing Phase: Nodes initially function like standard blockchain nodes, processing transactions and storing blocks locally. This ensures that the transaction throughput (TPS) is not negatively impacted during this phase.
  • Data-Reduction Phase: This phase, ideally performed during off-peak times or system backups (like Hyperledger Fabric's snapshot mechanism), is where the storage optimization occurs.
  • A Vector Commitment (VC) is calculated for the transactions within a block. VCs are cryptographic proofs that are significantly smaller than the actual data they represent, allowing for efficient integrity verification.
  • The block's hash and its VC are stored locally as an index.
  • The full block data is then transmitted to a remote DB server.
  • Finally, the block is deleted from the local node's storage.

When a query is received, nodes retrieve data from the remote DB server and use the stored VCs to verify its integrity. The system is structured into "Consensus Units" (CUs), where each CU manages its own remote DB server, preventing a single point of failure and reducing network overhead.

Datasets

Not discussed in the paper. The experiments utilize simulated blockchain networks and transaction loads.

Key Insights and Ideas

  • Decoupling Storage from Processing: The most significant insight is the separation of transaction processing from data storage. By allowing nodes to temporarily store blocks locally before offloading them, the system avoids the performance bottlenecks that can arise when immediate off-chain transfer is required.
  • Vector Commitments as Efficient Integrity Proofs: The use of VCs is crucial. They provide a compact way to verify data integrity without needing to store the entire block on every node, a key enabler for the massive storage reduction.
  • Snapshot Mechanism for Data Reduction: Leveraging Hyperledger Fabric's snapshot feature for the data-reduction phase is an elegant way to integrate the solution into an existing framework and manage the process during periods of low activity.
  • Consensus Units for Scalability: Dividing the network into CUs with dedicated remote DB servers helps manage storage and processing at a more granular level, contributing to overall scalability and resilience.

Key Results

  • Storage Reduction: SASLedger reduces the block file size on blockchain nodes by an impressive 93.3% compared to the original Hyperledger Fabric.
  • Performance Improvement: The advanced SASLedger solution enhances Transaction Per Second (TPS) by 9.6% compared to their previous solution, xFabLedger.
  • Query Latency: While retrieving data from the remote DB server is slower than from local storage, the overall query performance is manageable. In the worst-case scenario (all blocks in the remote DB), query latency increases by 31% for blocks and 67% for transactions compared to HLF.

Limitations and Future Work

  • Query Latency: Retrieving blocks and transactions from the remote DB server is inherently slower than accessing them locally. This leads to a decrease in query performance, especially when all data resides on the remote server.
  • Snapshot Dependency: The data-reduction phase relies on the snapshot operation of HLF. During snapshotting, transaction commitments are paused, which can temporarily impact performance.
  • Future Work: The authors plan to investigate mechanisms to address read-write conflicts when exporting local database contents during snapshotting. They aim to accelerate snapshot generation by allowing transaction commitments to continue during this process.

Summary updated

Abstract

Blockchain technology provides a “tamper-proof distributed ledger” for its users. Typically, to ensure the integrity and immutability of the transaction data, each node in a blockchain network retains a full copy of the ledger; however, this characteristic imposes an increasing storage burden upon each node with the accumulation of data. In this paper, an off-chain solution is introduced to relieve the storage burden of blockchain nodes while ensuring the integrity of the off-chain data. In our solution, an off-chain remote DB server stores the fully replicated data while the nodes only store the commitments of the data to verify whether the off-chain data are tampered with. To minimize the influence on performance, the nodes will store data locally at first and transfer it to the remote DB server when otherwise idle. Our solution also supports accessing all historical data for newly joined nodes through a snapshot mechanism. The solution is implemented based on the Hyperledger Fabric (HLF). Experiments show that our solution reduces the block data for blockchain nodes by 93.3% compared to the original HLF and that our advanced solution enhances the TPS by 9.6% compared to our primary solution.
BibTeX
@inproceedings{sun2021sasledger,
  author = {Haoli Sun and Bingfeng Pi and Jun Sun and Takeshi Miyamae and Masanobu Morinaga},
  title = {{SASLedger: A Secured, Accelerated Scalable Storage Solution for Distributed Ledger Systems}},
  booktitle = {Future},
  year = {2021},
  doi = {10.3390/fi13120310},
}
Figure