Ultrain CEO Guo Rui
1、 What is Blockchain 3.0?
The concept of Blockchain is on its way to becoming a topic for conversation in every household. However, I have asked myself many times, why there aren’t already real large-scale commercial Blockchain applications in place? Unfortunately, I can only tell you that, although Blockchain technology can create a completely new and disruptive business model, the underlying technological performance of the Blockchain public chain is not enough to support large-scale commercial applications. This is due to the technical bottlenecks which arise from inadequate system performance. Ethereum provides a good example. For all applications running on the whole network, the processing capacity can only reach around 20–25 transactions/second, and for an application that supports daily commercial operations, the minimum TPS peak requirement needs to reach around 2000–3000. The existing Blockchain 1.0 and Blockchain 2.0 systems, based on their performance statistics published by Bitcoin and Ethereum, are therefore unable to support large-scale commercial applications.
And why is the TPS minimum peak requirement so difficult to break through? Just as all distributed systems face the CAP problem when in the design phase, the Blockchain system also faces its very own impossible trinity: decentralization, security, and performance.
The challenge with decentralized designs is how to guarantee the level of decentralization of the network. This requires a network based on a peer-to-peer model. The machines in the network are all equal in status, with no node receiving special treatment or has any delegate voting power. At the same time, in order to ensure its decentralization, the network needs to be openly accessible , so that everyone can join , and it isn’t controlled by one or more “centers”
Another challenge related to the networks is security design. Specifically, how to ensure that the network is sufficiently secure and cannot be destroyed by a malicious node? In an open and network that’s directly linked to our economy, not only do potential participants buy machine rigs to join such networks, but potential malicious nodes will try to profit through network attacks. Then, how can we ensure the security of the network when participants joined with the intention to attack and break through the traditional security architecture? This is a challenge to the security design.
The performance design challenge is how to ensure optimal network performance with the lowest energy consumption.
Bitcoin, for example, was designed to establish a decentralized currency system. Therefore, it needed a completely decentralized open network, in order to ensure its security. To ensure a secure network, Bitcoin adopted the Proof of Work (POW) Mechanism, in which every node in the network needs to be mined through the calculation of hash value. Since every nodes participate in the consensus and record all corresponding data, making it costly for potential attacks . While Bitcoin guarantees the security of the network, it sacrifices the performance of the system. Currently, the operation and maintenance cost of the Bitcoin network reaches10 billion RMB per year, while the TPS performance is limited to 7 per second.
In order to solve this problem, BM wants to put forward its own ideas as early as 2014, which is a new consensus mechanism called DPOS. The core idea of the DPOS mechanism is to select a small number of nodes participating in the large-scale network of consensus based on the number of tokens held by each node. Once the consensus is achieved between these nodes, then an agreement is reached, which is different from Bitcoin, where blocks are formed. Taking DPOS’s latest project, EOS as an example. EOS selects 21 nodes as delegates nodes in each round, with one block formed within every 0.5 seconds, and confirms its formation within 3 minutes. According to the community’s latest EOS performance test report, its TPS is about 3,000 transactions per second. However, solutions such as this also have weaknesses. The most important weakness is two-fold. On the one hand, it is difficult, for the 21 nodes to resist large-scale DDOS attacks. Hackers can easily hack the 21 nodes to paralyze the entire network; on the other hand, the Blockchain emphasizes a completely decentralized and peer-to-peer network, where everyone in the network is treated equally and there is no specialization. Whereas, EOS design deviates from this concept, making the 21 nodes special. It is difficult for such structure to avoid problems such as corruption, collusion and centralization of control to the 21 nodes. Therefore, the industry is considering making EOS as a “semi-centralized” network. In the impossible triangle of the Blockchain, EOS is biased towards efficiency, at the expense of decentralization and security.
It is better to further improve the performance of the Blockchain system without sacrificing decentralization and security., which will inevitably require major innovations and breakthroughs in technology. This type of project is called the Blockchain 3.0 project. In recent years, especially since 2018, many teams around the world have made their own efforts, the most prominent of which is Dfinity, Oasis, Thunderlla, Algorand, Zillaqa, Ultrain. For these projects, we will analyze their principles and characteristics one by one.
2、 Analysis of Top Tier Blockchain 3.0 Projects
Dfinity：Threshold Signature Scheme — A New Direction for Consensus
This project launched in Silicon Valley in 2016. Its founder, Dominic is one of the core members of the Ethereum Early Cryptography Association. Dfinity’s most prominent contribution to the crypto industry has been its introduction of threshold signature technology into consensus algorithms. The consensus idea of Dfinity is as follows: First, the nodes of the whole network are randomly divided into various numbers of groups (N Groups). At the beginning of each round of consensus, block and notary nodes are randomly selected. Upon the formation of a block, the notary nodes choose which one will be the final block. At the same time, of the notary node will randomly select nodes that forms a block, and complete consensus to continue its function. However, since there is no power consumption as a measurement of security prevention (similar to POW mechanism), Dfinity’s generated block and notary nodes in groups. For example, each nodes that forms block is based on 400 different machine that acts as nodes to build blocks. The threshold signature technology is used to guarantees the integrity of the data block. After more than 51% (more than 201 machines) are randomly selected and execute their individual signature, Dfinity can then generate data blocks verified by the third party and can generate the random number of groups for the next round. This process greatly improves the overall security of the network. This system turns an original attack on a single node into an attack of a group of nodes, which becomes much more challenging, thus, overall system security is improved. In the field of cryptography, VRF (verifiable random functions) is a crucial component. Threshold signature function is another key component. However, the biggest breakthrough of Dfinity is the fact that its technology incorporates both VRF and threshold signature, which guarantees that its algorithm can be implemented. Based on this consensus, Dfinity claims that it has reached hundreds of transaction per second, and the confirmation time for each block is 7.5 seconds.
Dfinity’s consensus mechanism is well designed, but there are still flaws remaining in the system. Because the establishment process of the “group” in the threshold signature is very complicated, each group must be kept for a long time after being established. At this time, there is a potential game theory problem. That is, the group signature can be predicted through the collusion of multiple members, the cost of which is very low. Each group member already knows which group they belong to, through various means such as the Internet. This enables them to easily find members of the same group to collude with. Colluding members can collaboratively acquire the group’s private key through calculation and quickly predict the next round of random numbers, thus undermining the integrity of the network. Because such attacks are very difficult to find, they can be achieved at no cost. We have already submitted this question to the Dfinity team and have not yet received a response.
Algorand：Consensus Based on Random Calculation with Impressive TPS
Algorand is the Blockchain Consensus Agreement released in May, 2017 by the Turing Award winner and MIT Professor Sivio Micali. The main idea here is to combine the random selection and BFT algorithms to achieve high TPS on a completely decentralized network.
The first step is the role confirmation phase. For a large-scale network, each node starts with a VRF (verifiable random function), which generates a voucher. The node randomly selected with each voucher participates in the consensus and is called “the voter”. The one with the smallest voucher value is selected as the “proposer”.
The second step is the consensus grading phase. In this step, the proposer is responsible for assembling the candidate blocks. The voters then agree upon the leader node for the current round. They are also responsible for confirming the potential blocks
The third step is the binary Byzantine phase. In this phase, the verified voters will vote to either accept (assuming there is no problem with block) or not accept (it is deemed to have an error such as double spending) the candidate block while replacing the defective block with an empty one)
The final step is to broadcast data to the entire network: Algorand effectively improves the system’s TPS on the basis of ensuring network security by randomly selecting the consensus nodes for every round. According to the data in the paper, the consensus transaction is 750 Mbytes per hour. According to Bitcoin, each transaction length starts from 250 bytes. Thus the calculation would be 75010241024/60/60/250=873.8 TPS;
As we introduced and summarized the details of the Algorand algorithm above, let’s now discuss how there is still room for improvement.
Algorand’s main function now is to transfer transactions. Its function is similar to that of Bitcoin, and as the most important smart contract improvement in Blockchain 2.0, it does not address how to support smart contracts within its network.
The Algorand network operates under the assumption that the amount of “honest” voter nodes are over 2/3, but does not mention how it can guarantee the honesty and integrity of the network.
Algorand’s algorithm requires completely random selection at each step of processing. Frankly speaking, the overall complexity of the project remains high. After the paper was published, Algorand has organized a team to quickly promote the development of the project, and the community has been waiting for Algorand’s engineering establishment
Thunderlla：Combining the POW and the POS
The founder of Thunderlla is Elaine Shi, a professor of computer science at Cornell University. Thunderlla proposes a new algorithm, assuming the accelerator node and over 3/4 of the committee nodes in the network are “honest” and proper, while the network is functioning well, Thunderlla can implement fast and asynchronous processing with a confirmation time of <1 second, and can process all transactions almost instantly. And when there is an abnormality in the network, such as the emergence of the Byzantine failure, the network will start the cool-down mechanism and switch to the traditional Blockchain consensus (with slower processing performance). This cool-down mechanism ensures network security and sustainability, while restoring the system. Once restoration is completed, it automatically switches back to the original mode. Therefore, during optimal functioning, the network exceeds the processing speed of the current Blockchain by over 1000 times, and when potential issues emerge, the network can still resist attack from the 49% malicious nodes through the traditional (slow) chain approach. For the traditional chain approach, Thunderlla can operate on Bitcoin, Ethereum, or any other Blockchains. In other words, it, instantly creates a friendly, safe and reliable atmosphere in a malicious environment.
This algorithm can be seen as a mixture of POS and POW, hoping to attain the advantages of both POW and POS algorithms, but there are several key issues that remain questionable. For example, Thunderlla never explicityly mentions how it can ensure that ¾ of its committee member nodes are honest and how it will remove malicious committee members. For example how do we select the next accelerator, In terms of progress, Thunderlla has yet to announce any further progress after the publication of the initial paper.
Ekiden of the Oasis Lab：Performance Improvement Based on a Trusted Environment
The project was launched in 2018 by its founder, Dawn Song, who is an associate professor of computer science at the University of California at Berkeley. Ekiden’s main idea is to separate the consensus layer from the computing layer. In the computing layer, the hardware is composed of TEE (Trusted Execution Environment), such as Intel’s SGX, The calculation of the smart contract is executed through Tee, and the consensus layer uses POW or POS to verify the result of TEE calculations. This method has two characteristics:
The calculation node and the consensus node are separated, and the computational node can execute with arbitrarily complex logic. The calculation result is mutually verified by a small number of trusted computing nodes, so the execution efficiency is high, basically as efficient as executing on a single machine. The network also supports execution of multiple contracts simultaneously with different machines
Privacy protection: Only the encrypted data (even the encrypted contract code) is stored on the chain. The decryption is only done in the TEE, and then the result is calculated, encrypted and returned to the chain.
From the overall design of Ekiden, its security relies entirely on the TEE’s trusted execution environment. Although it solves the security problem ingeniously, there are security risks remaining. In March, 2017, researchers from Graz University of Technology in Austria cracked the protection of SGX. Therefore, it is still important to check whether the security of a single unit of hardware is reliable. At the same time, the principle of TEE is that the private key is stored with the chip manufacturer. For example, for SGX, the security of the private key is Intel’s responsible, such arrangement allows the chip manufacturer to become a center of nodes. This contradict to the original intention of Blockchain of creating a complete decentralized network
Ekiden’s progress is quite impressive. The website has already allowed users to submit their own machine configurations to test the network application. Unfortunately, there is no data on the performance of the test network.
Zilliqa：TPS Performance Based on Fragmentation
Zilliqa is the first public-chain project with a test network that incorporates segmentation technology, and its function, which is only to provide transfer transactions, is relatively simple. The core idea is to greatly enhance TPS through fragmentation. We can imagine a traditional Blockchain main chain as a single-core CPU, which can only process a limited number of calculations. If the main chain is composed of multiple sub-chains, think of it as a multi-core CPU processor whose performance can achieve multiple levels of improvement. However, the algorithm still has the following shortcomings:
Zilliqa data processing: Data storage itself is not fragmented. Improving performance in data calculation will bring efficiency problems in data storage/synchronization/sharing, etc., which will greatly restrict its performance. At the same time, the implementation of the upper-level smart contract will be extremely unfriendly. This method is more suitable for specific application scenarios, such as scientific computing. But it is very difficult for developers to use.
Cross-sharding is not formed: There is no detailed description of how to manage global sharding: the most important problem with sharding is how to deal with cross-sharding, and Zilliqa has no solution yet.
At present, Zilliqa has already launched a test network, and the testing performance statistics of the internal network’s is based on the operation of 3,600 machines, with 6 sharding slices, and reach 2000 transactions per second
3、 Next Generation Blockchain 3.0 Project Ultrain
Above, we introduced some of the most recent Blockchain 3.0 projects. Now, as we try to break through the Blockchain impossible triangle, Ultrain also is making its own contribution with its proposal of a new R-POS consensus. Let’s use a metaphor as an example. Similar to the DPOS consensus mechanism adopted by EOS, a large number of people vote for a fixed number of groups to make decisions based on the number of tokens held by the group. The R-POS mechanism randomly selects a few people and makes its own proposals among a group of people. It then randomly selects 10 times more people to verify whether the previous proposal is correct. Ultimately, the proposal will be correct. The correct proposals are then integrated to form the final decision.
Therefore, the core idea of R-POS is to change the selection method of participating in the consensus node from the entrusted election to a random selection process on the basis of DPOS. This ensures that everyone on the entire network has an equal chance to be elected to participate in each round. Consensus nodes not only protect the decentralization of the network, but also greatly improve performance. At the same time, Ultrain guarantees R-POS by introducing parallel technology, fragmentation technology, device fingerprint technology and multiple cryptography improvements. In terms of further improvements in security and performance, the following is a brief introduction to the R-POS consensus process:
Each Round of R-POS consensus mechanism can be separated into four different stages. Each round of consensus confirmation takes 10 seconds:
The first stage is the role confirmation stage. Here, nodes are randomly elected from the entire Ultrain network. And each node’s role is determined in the current round of consensus by applying the VRF function. In each round of consensus, the roles of nodes are divided into three types:
Block formation node: responsible for assembling the potential block in the current round; node will be selected as the formation node in each round;
Voting node: responsible for the next stage of voting and to confirm the identity of the formation node of this round;
Data recording node: does not participate in block formation, and provides data recording service after the block is determined;
The probability of any single node to be selected as a block formation node or a voting node depends on the input parameters of the VRF. These parameters are based on the number of Tokens held by the node, the performance of the devices (node) and the credibility of the node. The higher these three parameters are, the higher the probability of selection. Device credibility; for each node a hardware fingerprint will be generated that cannot be tampered with. The credibility of the node and each device is based on this hardware fingerprint. Nodes that consistently show good behavior are more likely to be selected.
In short, Ultrain first increases the difficulty of external and internal attacks by randomly selecting nodes. Secondly, based on the credibility of device fingerprints, it establishes a model for the credibility of a single node, which makes attacks more difficult to execute. The device needs to accumulate its credibility score to increase the probability of being selected, and this takes time. This is similar to the notion of using time consumption as a replacement for computing power consumed by hash mining, while the system provides better Security; Finally, based on the token-locking mechanism, the cost of malicious attack increases, therefore improving the overall security of the Ultrain network in many ways.
In the second stage, the parallel consensus stage , selected formation nodes respectively assemble the candidate blocks of the current round in parallel. Then, the voting node reaches a consensus on the formation nodes to determine if the block is received by most of the nodes; Here, Ultrain has changed from the original one formation node to multiple block formation nodes to work in parallel, which greatly improves the system’s TPS. For multiple block formation to work at the same time, even if a block node is attacked, as long as one node reaches a consensus, it can ensure that the system does not generate an empty block, thereby improving the activity of the system, and achieving the core of the target. The technical challenge lies within the parallel BA algorithm, and thus may cause a large number of “network storms”. To solve this problem, Ultrain introduces redundant coding technology to divide the message into multiple transmissions, ensuring that the largest message is broadcast with limited network bandwidth. Quantity, optimizes network throughput. Ultrain also introduces threshold encryption technology to allow candidate blocks to be delivered in small chunks during delivery. That is, each node is guaranteed to be unable to receive enough messages, know the content of the messages, avoid the tendency to pass consensus messages, and improve fairness;
The third stage is the parallel BA stage: Once again randomly selected voting nodes will agree on the candidate blocks in the parallel consensus stage, ensure that everyone accepts the same block combination, and the block based on accepted consensus will eventually form blocks based on certain cost
The fourth stage is broadcasting the blocks identified during this round to the whole network, completing the current round of consensus;
We illustrated the operation of the consensus in the case of “single-slicing”. In order to improve the efficiency of the consensus, Ultrain also introduces fragmentation technology. Ultrain’s sharding technology takes into account both data and data storage sharding. The core idea is that when processing across shards, the data needs to be stored across them redundantly. That is, by space for time. This results in overall improvement of consensus efficiency.
In general, the core technology of the Ultrain Consensus is VRF+BFT. The introduction of parallel technology and fragmentation technology greatly improves the performance of the system, expands the security of the system through device credibility, and achieves the maintenance level of POW to further improvement of system performance.In addition, Ultrain has released the concept network of R-POS in early July, and we deployed the Ultrain system to 1000 nodes on the public Amazon cloud. After actual testing, the network can reach an average of 3000 TPS with a confirmation time of 10 seconds. Its performance far exceeds the existing Blockchain 3.0 project. In addition, it is expected that the sharding technology will be deployed simultaneously after the publication of Ultrain’s network in April of next year. At that time, the TPS of the network can be improved by a factor of several dozen times.
The landing of the Blockchain 3.0 project is the only way for Blockchain to empower the real economy. It is hoped that all the public chain teams around the world can develop rapidly and solve the core issue, the fact that the existing Blockchain public chain cannot support large-scale commercial applications. Let Blockchain technology help the real economy to establish a new business model and substantially improve social productivity.