Building the First Real-Time Blockchain - TOKEN2049 Dubai 2024

Yes. Has anyone heard of Mega East? No one. That's why I'm here. Um, cool.

So, my name is Shu Yao So, my name is Shu Yao . I am one of the co-founders of Mega East. We are a group of passionate developers pushing the boundaries of what a fast, real-time blockchain can be. So, if there's one thing you take away from our conversation today, it's that Mega East is taking a radical approach to scaling blockchain. We are here to fundamentally open up new uses and new experiences.

We are ready. OK. Let me tell you a little about my team. We are a team of brilliant researchers and business professionals business professionals . I have two technical co-founders.

Elon, a graduate of Stanford's PhD program in computer science . He specialized in low- latency data center computing. My second technical co-founder, Lei, graduated from MIT in computer science. He specialized in high-bandwidth networks . Does anyone here know EigenLayer?

EigenLayer. There is a well-known white paper on data accessibility, which EigenLayer is famous for. My co-founder Lei is its lead author. He wrote it while he was at MIT. And I am Shu Yao.

For the past many years, I have led business development at ConsenSys. I studied at Harvard Business Harvard Business School. Let's be honest: blockchain has not yet scaled. Everyone jokes about TPS. TPS is an outdated meme; You can claim 1 million TPS, but in reality no blockchain has yet scaled beyond a few thousand TPS.

We've all seen meme coins appear on some blockchains and the network just crashes. So why is that? The problem is not the equipment. In fact, if you look at the hardware components in a blockchain— blockchain— computing power, disk I/O, and network bandwidth—each of them is capable of processing 100,000 TPS. So why haven't we scaled blockchain properly yet, if it's not a hardware issue?

What is the difference between potential and reality? Our short answer: software inefficiency. If you look at a modern blockchain, it has two important components. Consensus and execution. Consensus execution.

Consensus is essentially the ordering of user transactions and their consistency. A transaction is the execution of these ordered orders. Execution orders. Execution is the execution of these transactions in order. Both components are extremely important for the blockchain, and they determine performance: it depends on the slowest of them.

Let me give you an example. It rained in Dubai, didn't it? Has anyone been hurt by the rain? Everyone. Imagine it rains in Dubai again, God forbid, and a sewage system with two pipes: a larger one and a smaller one.

How much wastewater do you think such a pipe can pass per second? What does it depend on? From a smaller pipe, right? Because the smaller pipe is a bottleneck for the TPS. The same with blockchain.

You have consensus and you have execution. What runs slower actually determines the performance of the blockchain. We figured it out here. So, based on this, why why doesn't layer 1 scale? Layer one fundamentally cannot scale because layer one nodes must execute all transactions simultaneously transactions simultaneously .

So the first level nodes must reach consensus. They must ensure data availability. They must also perform the transactions themselves. When you are forced to do so many things at once, you are essentially duplicating your efforts. When you duplicate efforts, you become very slow.

I will give another example. Let's say I'm a teacher Let's say I'm a teacher . I have a thousand students. Some of my students are probably very talented. Some students probably need a little more attention.

But the speed of my learning actually depends on those students who are slow. And this is exactly what the modern, first-level blockchain faces. It can't work fast because there will always be a node that is simply slow. Therefore, level one has fundamental bottlenecks and is fundamentally inefficient. Let's look at layer 2.

Are they better? To some extent yes, but not entirely. Level two solved one problem—the one problem—the consensus level. People are familiar with optimistic rollups and ZK-rollups. Yes.

No. This is what they did This is what they did , resolving the consensus issue. But none of the existing level twos have even touched the execution level. So, going back to my example of the Dubai rains, it's like there are two small drainpipes on level two, and they fixed one of them. So now you have one large pipe and one smaller one.

What is the result? You still haven't solved the scalability problem, have you ? Because you still have a smaller pipe. But we believe that level two has the most unique and underexplored opportunity for scaling, as one pipe has already been fixed. Now already been fixed.

Now , if you fix the second pipe, you get a super high-performance blockchain. And that's where Mega E comes in. We aim to solve the last mile scaling problem, which is the level of execution. Our goal is to achieve over 100,000 TPS and latency of less than one millisecond. Realize this.

I don't think anyone has seen a blockchain that works in a similar way. So how do we do it? We have two special secret ingredients secret ingredients . The first is the specialization of nodes the specialization of nodes . What is node specialization node specialization ?

This is the idea of separating execution from a full-fledged second-level node. So, you are effectively offloading heavy computations to a smaller group of execution nodes. In my original example original example : you have a thousand students, and instead of teaching the able and the less able at the same time, you isolate the latter and give them special attention so they can move quickly. And this is precisely the essence of node specialization node specialization . You focus on fewer nodes to maximize their power, and then have a multitude of other decentralized nodes to check the execution nodes to make sure they are not cheating and keep the blockchain running.

Our second secret ingredient is made up of many components many components . And the general idea is that by having fewer sequencers, you can optimize their performance as much as possible. There are several optimization methods. Today I will name four. The first is Merkle tree support.

Maintaining a Merkle tree consumes 90% of CPU cycles, and we use new data structures to solve this problem. The second is the overhead costs of implementation. We implement JIT We implement JIT compilation to reduce this overhead. The third is in- memory computing. For those who are not technical: loading data into memory is the fastest storage layer for everything.

And the last one everything. And the last one is parallelization. Who here is familiar with parallelization? So, parallelization doesn't really scale the blockchain. The maximum that parallelization can provide parallelization can provide is perhaps a two- or three-fold increase in performance.

And the reason is that not everything can be parallelized. For MegaETH, parallelization MegaETH, parallelization is a tiny part of our technical stack. But I always feel obligated to mention it because everyone is so into parallelization. However, I think that soon people will understand: parallelization, especially at the first level, will not solve anything. So what does this mean?

Isn't that right? Remember, the one thing I want everyone to take away from this presentation is that MegaETH takes an extreme approach to scaling the network, and by combining centralized block creation with decentralized validation, you actually open up a new set of possibilities for users and new use cases. So, CLOB, right? Centralized limit order book. She didn't work on any blockchain.

Didn't work on EVM, didn't work on Solana. I don't know where she worked, but she will work for MegaETH. Live sports betting : who watches Love Island? Imagine, you watch this, guy, you keep raising your hand, thank you for your support, even though you're not even on my team, that's great. Imagine you're watching Love Island and you can place real-time bets with your friend on the outcome.

So that's pretty cool. This is an optimized blockchain. Imagine a completely on-chain game, right? Never have to worry about the engine again. Developers can safely host everything online instead of worrying about gas optimization.

And that's where the potential lies. And lastly: I think there are many use cases that our team, to be honest , hasn't even thought of. Just like when the internet became mature, who would have thought that YouTube would appear? So for me, Mega is the technology that will open up the next YouTube in Web 3. We are here to make Ethereum great again.

Join us in development. Thank you.