Next-Gen Blockchain: Scaling PoW-based Public Chains with Tree Graph
On June 20th, Y-Talk hosted “Next Generation Blockchain” event in the Olympics Sports Center, Beijing. The event mainly focused on next-generation blockchains and invited speakers from bloXroute Lab, QuarkChain, ArcBlock and YeeCo. Conflux’s Co-Founder Fan Long was also invited to participate, and gave a presentation on optimizing throughput and confirmation latency of public chains.
Fan Long started by addressing the bottleneck issue which prevents scalability in current public chains like Bitcoin. He observed that the bottleneck issue lies not in the hardware capabilities, but within the consensus mechanism itself — specifically, the inability for current public chains to process non-conflicting transaction blocks concurrently. The forks inevitably formed due to bifurcation and high concurrent networks in Bitcoin, for example, is wasteful and inefficient.
We aim to turn that ‘waste of resources’ into a solution that will enable chain scalability.
Conflux organizes blocks into a novel structure we call the Tree Graph, which allows each block to have two edges: a Parent Edge (connected to its parent block) and a Reference Edge (which corresponds to blocks generated before the current block).
The block order is determined by our GHAST algorithm, which assigns each block a weight, starts counting from the Genesis block and iteratively advance to the child block with the largest subtree. With GHAST, each block has different adaptive weight and the chain is secured against all attacks — including double-spend attacks. In our testnet, each block generated with GHAST takes 0.25 seconds, and transaction confirmation takes 23 seconds.
Evaluations and Results
Fan Long briefly brushed upon the architecture of Conflux before wrapping up with our latest evaluations and test results.
On Amazon’s EC2 clusters, Conflux ran 12k full nodes and reached 9.38 Mbps, while each transaction confirmation took an average of 23 seconds.
When we ran 400 full nodes to process Ethereum’s past transactions and payments, we found that 7 months of transactions in Ethereum took 40 minutes in Conflux, with a confirmation time averaging 17 seconds and a speed of 1400–3500 transactions per second.
Our results indicate that when running Conflux, the consensus protocol is no longer the throughput bottleneck. Instead, the bottleneck is at the processing capability of individual nodes.
During the brief Q&A session after, when asked about the main difference between Conflux and Algorand, Fan Long iterated that whilst there are specific technical differences between the two, our main difference lies in our vision.
As our Co-Founder Forgiven Zhang said: “Conflux is a team made by the people, for the people.” Conflux is committed to developing a decentralized public chain that remains open to anyone to join right from the beginning of its conception.
This vision is what drives our team to innovate, rather than sacrifice any of the three major points of blockchain — decentralization, confidentiality, and security.
Our view of the future
As explorers of blockchain technology, we believe that in the future blockchain will revolutionize payment, logistics, entertainment, identification, and beyond. Our aim is to be part of this revolutionary movement and advance the development of decentralized finance, intellectual property, traceability, sharing economy and internet — and create extended possibilities for the evolution of society.
— — — — — — —
Conflux, where valuable bits are exchanged and validated.
Our team is full of the best and brightest (Alan Turing Award winner? Graduates from top universities like MIT, UC Berkeley, and Tsinghua University? National Olympiad winners? We got them all). Get to know us more on our Telegram!