We want to express appreciation for being able to take advantage of the work that’s been done by others before us. We didn’t invent the language or mathematics. Everything we do depends on other members of our species and the shoulders that we stand on. We try to use the talents we do have to add something to that flow.
— Steve Jobs
epic is a POW (proof-of-work) chain that achieves an order of magnitude improvement on the throughput over current mainstream cryptocurrencies, e.g, Bitcoin, without compromising security and decentralization. The designed max TPS (transaction per second) is 1024 per second. Miners collectively create 8 blocks per second on average, each containing up to 128 transactions. You may join the epic network by running a node following the instruction. epic uses the same UTXO system as Bitcoin, consequently no rosy applications like Hello Kitty or Texas Poker for now.
The main idea is to expand a chain of blocks, e.g. the Bitcoin blockchain, to a more general graph of blocks organized in a certain structure. Refer to our paper for the full description or this video presentation for a verbal explanation. Here we provide a brief description and some intuition. The blocks are organized in a structured Directed Acyclic Graph (sDAG):
- Each miner forms a
peer chainby connecting his newly mined block to his most recently created block. Miners, depending on their mining power, has the flexibility of creating several parallelpeer chainsby splitting their mining power to these parallel chains. - Each block turns out to be either a
milestonewith probability$$p$$ , or aregularblock with probability$$1-p$$ .$$p$$ is determined by the current number of transactions getting processed and the hashing power of the entire network. In general,$$p$$ increases as the transactions pile up and as the network hashing power grows. Before a block is successfully mined, the miner has no idea which type it will turn out to be. - In addition to connecting to its immediate preceding block in the peer chain, each block, regardless of its type, must point to two additional blocks: the most recent
milestoneand anotherregularblock on a differentpeer chain.
In our sDAG, all the milestones form a chain, connecting the parallel peer chains. The connectivity among the parallel peer chains is quite strong since each block needs to point to another regular block on a different peer chain.
Organized in the sDAG, we will be able to break a big block, e.g., Bitcoin block containing thousands of transactions, into smaller ones, allowing quick dispersions of information across a peer-to-peer network continuously. The motivation for such a design is that the state-of-art Bitcoin blockchain network propagates a block of 1 Megabyte periodically every 10 minutes on average. During the interval, the capacity of the network is wasted. Many smaller blocks propagating over a peer-to-peer network in a more "continuous" fashion significantly improve the utilization of the network capacity. Fundamentally, to build a high TPS system, the corresponding amount of information has to be synchronized across the network. Increasing the utilization of network, before the current Internet infrastructure upgrades to be faster, is important.
While benefitting from the increased utilization of the network capacity, many smaller blocks organized in a more general structure than a chain certainly brings challenges on reaching consensus. This is resolved by the design of our sDAG. We say a block is confirmed by a milestone if it is connected to the milestone directly or indirectly. For each milestone, it is associated with a level set, all the regular blocks confirmed by this milestone but not by any preceding milestone. Essentially, all the level sets form a chain, same as the Bitcoin blockchain. In this way, a consensus is reached in the same way as the Bitcoin by POW and Nakamoto consensus proposed in the original Bitcoin white paper. As a result, the security of epic is guaranteed in the exactly the same way as the Bitcoin.
To ensure a certain security level, Bitcoin needs to wait for a certain number of blocks to confirm a transaction. For example, to limit the probability of a successful attack by 0.001 given 10% malicious mining power, we will have to wait for 6 blocks. In order to achieve the same security, we need to wait for the same number of milestones provided the same percentage of malicious mining power. A simple calculation based on the above parameters shows that a milestone will be created on average every 16 seconds in epic. We can achieve a faster transaction confirmation because the speed of creating milestones is faster (16 seconds v.s. 10 minutes), not because we are using a more superior consensus mechanism. However, we are working on incorporating the idea of multiple verifying chains proposed in the paper so that confirmation can be much faster because the number of milestones we have to wait is reduced.
In a high TPS system, different miners may work on the same transaction from the mempool causing waste of capacity, despite that consensus is not affected. To reduce waste, we design the transaction assignment rule. A miner can process a transaction only if the hash of this transaction and his most recent block exhibit a certain pattern. Intuitively, transactions in the mempool are "distributed" to miners (more precisely peer chains) in a probabilistic fashion. Our paper shows a quantitative computation of how such probability affects the waiting time of a transaction in the mempool and the wasted capacity. Under our carefully chosen parameters, the capacity waste is controlled under 1% and the waiting time in mempool is roughly in the order of seconds when the system is underloaded. In fact, the waiting time will largely depend on the speed at which all the users generate transactions if that speed is close to the system capacity, namely the max TPS.
Blocks are created by POW, using the cuckoo algorithm. Users may choose to provide transaction fees in the same way as in the Bitcoin system. In addition to this reward, each block brings a block reward. Each milestone brings an additional reward depending on the number of regular blocks in its level set. Such a reward scheme incentivize miners to connect to recent blocks on another peer chain to enhance the connectivity of the sDAG.
In Bitcoin, each block reward generates a coinbase. If blocks are mined at a much higher speed, the number of UTXO will explode. To avoid this, we let miner redeem his reward by creating a normal transaction every once in a while, with the frequency completely depending on the miner.