[Paper Review] Blockclique: scaling blockchains through transaction sharding in a multithreaded block graph

Sébastien Forestier, Vodenicarevic, Damir+1 more|arXiv (Cornell University)|Mar 24, 2018

Blockchain Technology Applications and Security22 references15 citations

TL;DR

Blockclique proposes a multithreaded block DAG architecture that scales blockchain throughput by sharding transactions across T parallel threads, where each block references one parent from each thread. By ensuring transaction compatibility through input-address-based sharding and extending Nakamoto consensus, the protocol achieves over 10,000 transactions per second with low stale rates in realistic network conditions.

ABSTRACT

Decentralized crypto-currencies based on the blockchain architecture under-utilize available network bandwidth, making them unable to scale to thousands of transactions per second. We define the Blockclique architecture, that addresses this limitation by sharding transactions in a block graph with a fixed number of threads. The architecture allows the creation of intrinsically compatible blocks in parallel, where each block references one previous block of each thread. The consistency of the Blockclique protocol is formally established in presence of attackers. An experimental evaluation of the architecture's performance in large realistic networks demonstrates an efficient use of available bandwidth and a throughput of thousands of transactions per second.

Motivation & Objective

• Address the inefficiency of traditional blockchains that underutilize network bandwidth, limiting throughput to ~5 tx/s despite available capacity.
• Overcome the scalability bottleneck of single-chain blockchains by enabling parallel block creation without sacrificing consistency or security.
• Design a scalable, decentralized blockchain architecture that maintains full node verification of all transactions while enabling high throughput through transaction sharding.
• Ensure security against up to 45% adversarial resource control by formally proving protocol consistency under adversarial conditions.
• Demonstrate that a natural extension of Nakamoto consensus—via multithreaded block graphs—can efficiently utilize network bandwidth and achieve high transaction throughput in large-scale networks.

Proposed method

• Introduce a multithreaded block DAG data structure where each block references one parent block from each of T threads, ensuring global consistency through cross-thread dependencies.
• Implement transaction sharding based on input address assignment: transactions are assigned to a thread based on the address of their input UTXO, ensuring that blocks in different threads are inherently compatible.
• Extend Nakamoto consensus to the multithreaded DAG by preferring the chain with the highest cumulative work across all threads, preserving security and finality properties.
• Use a fixed block creation interval t₀ and a consensus bitrate C_B to model network performance, with block size S_B determined by C_B and t₀.
• Ensure all nodes verify and process all blocks across all threads, eliminating the need for cross-shard communication beyond parent block references.
• Formally prove protocol consistency and security under adversarial conditions, showing resilience up to 45% of total network resources controlled by attackers.

Experimental results

Research questions

• RQ1 Can a blockchain architecture achieve high transaction throughput by enabling parallel block creation across multiple threads without compromising consistency?
• RQ2 How can transaction sharding be designed such that blocks created in parallel are inherently compatible, eliminating double-spending risks?
• RQ3 To what extent can network bandwidth be efficiently utilized in a decentralized blockchain by structuring blocks as a multithreaded DAG with cross-thread parent references?
• RQ4 What is the maximum achievable consensus bitrate and transaction throughput under realistic network conditions (e.g., latency, bandwidth, node count) with the proposed architecture?
• RQ5 How does the protocol maintain security and consistency when up to 45% of network resources are controlled by an adversary?

Key findings

• The Blockclique architecture achieves a transaction throughput exceeding 10,000 transactions per second in large-scale network simulations with realistic bandwidth and latency.
• With T=32 threads, t₀=32 seconds, and C_B=12 Mb/s, the protocol achieves a confirmation time of 36 seconds for t₀=16 seconds, decreasing to 142 seconds for t₀=64 seconds.
• Even with 30% block loss (μ=0.3), the throughput remains high at 8,070 tx/s, demonstrating robustness to network losses.
• The protocol maintains a stale rate below 1% in networks with up to 4,096 nodes when C_B is set to 12 Mb/s, indicating stable consensus under high load.
• In low-bandwidth conditions (B=4 Mb/s), the architecture still supports a consensus bitrate of up to 2 Mb/s, yielding approximately 2,000 tx/s.
• The architecture efficiently uses available network resources, achieving a consensus bitrate close to the theoretical limit of node bandwidth, even in large and high-latency networks.