On November 5, PKUSSM-OCTA innovation laboratory carried out an academic salon activity under the theme of ’Discussing Trias consensus technology’. Dr. Ruan Anbang (Trias founder and CEO), Dr. Wei Ming (Trias CTO), professor Shen Qingni of Peking University, and several doctoral students in the laboratory discussed the implementation status and technical details of Trias project, and further exchanged on the future innovation direction of Trias project.
To help you better understand the salon, here’s a part of Dr. Wei’s speech.
Today, we mainly discuss consensus from the four aspects of birth, ladder, communication and change. Here I would like to tell you that if you choose decentralization, you have to get rid of the original ideas. Because if you mention those ideas in technology, such as consensus algorithms, someone must think you are not engaged in blockchain.
I also want to emphasize that in traditional systems without TEE technology, people choose to hide from others. But in Trias system, we use the challenger model. To select the most trusted nodes in the entire network, you need to be considered the most trusted by all the nodes.
Our team used to joke that the idea was a bit like the Dark Forest Rule of The Three-Body Problem. The author of the novel argues that no one in the world wants to be discovered by a more advanced civilization. However, in the world of Trias, if you want to be a supernode, you have to be the strongest one and be able to be challenged by all the nodes in the entire network.
At present, no other blockchain has selected supernodes in the same way like us. EOS, for example, selects supernodes for votes. Trias needs to confirm through TEE whether the nodes have the ability to become supernodes and be the most secure node in the entire network.
Currently, Trias’s challenger model is unique in the blockchain industry. Of course, we still have many problems to solve.
Currently, our first vision of the state diagram has been developed according to the PPT, but there are still some bugs need to be improved continuously.
Let me introduce the process, the first for any node that wants to join Trias, you must broadcast to the entire network and let everyone know your address. You need to let all the nodes know through Gossip and P2P that you want to be the best node. Next, all information about your machine will be sent out, such as IP address, hash values representing trusted states, and the machine’s current trusted white list. Other nodes that get this information become challengers.
Because when spread through Gossip, it actually jumps like the small-world network. When it’s jumping, you can’t let everyone in the world know every second.
So, we’re actually judging by two elements, one is network state, the other is timestamp. Let’s say I know through UTC that it’s 12:00 and then I suddenly get a challenge from someone who wants me to challenge him at 11:57, and it’s been three minutes past, which means we’re far apart.
From the state of the network, we are certainly not a small-world network, because we are not close. So, if I get a challenge, it was sent out a year ago, so it doesn’t make sense. We have to be at a very close distance. Therefore, we rely on Gossip to determine if the other belongs to the small-world network.
Each machine’s default trusted nodes need to pass the challenge, which means they both challenge and accept the challenge.
Why are there blue nodes in the figure above, for fear of independent local authority. If we simply rely on Gossip to determine this distance, that is, through Internet communication, dictatorship will appear. Because in a certain region, it’s always these seven nodes, and if you don’t change its network state, it’s always the first one.
So, if you run them on Trias for a year, they’re likely to form small teams that are centralized. To solve this problem, we will send the supernode. That said, it’s not enough to rely on Gossip alone, but to send our superdelegates to be part of the board. So, in a team like this, they’re constantly confirming each other. Once we realize that there’s a node, no matter how many times it’s been right before, even if it hasn’t been wrong in 10 years. If the hash value calculation is wrong at some point in time, we’ll kick it out immediately.
Then, we’ll find network nodes closer to the team through Gossip and add them in. Besides the distance, the value of TEE should also be considered. The higher the value of TEE proves that you have basically made no mistakes in the entire history, and the better your credibility.
