Taking an abstract view, many modern economic activities are essentially the flow of information and value, and the conversion between them. In the past 100 years, the transfer of information has evolved from discrete events through pigeon networks to continuous flow through the speed-of-light Internet.
However, the “value” transfer part is clearly lagging behind: it is far from “speed of light” and is still very much discrete events between segregated financial silos defined by institutions and worse, heterogeneous interpretations of contractual logics. This mismatch creates a devastating bottleneck in human evolution: no matter how fast information flows, the segregated and slow value transaction is limiting the productive conversion between value and information.
We need speed-of-light (i.e. “cSpeed”) value transfer and state transitions.
Then comes blockchain technology with incentive compatible distributed consensus. This revolutionary abstraction of trust among untrusted parties brings the foundation to break down the segregated financial silos and drastically expand the scope and accessibility of global value flows. However, despite the promising potentials, blockchain in practice is deviating further away from the cSpeed vision due to its low processing power when compared to traditional value transfer tools. Scalability is a fundamental challenge that is blocking the path towards mass adoption of blockchain technology.
Before we dive into some grand solution for scalability, let’s first understand the problem space better by answering two basic questions:
How fast does blockchain actually need to be?
To figure out this “ultimate goal”, we start with the above-mentioned principle that the speed of value transfer should match the speed of information transfer. In just one second, the Internet can process roughly 2M emails, 65K Google Searches, 72K YouTube video plays happening. These are just a few examples of Internet’s applications and the total number easily amounts to billions of “transactions per second (TPS)”. It is not to say that we need to bring these applications to the blockchain, but to get an understanding of the order of magnitude needed for blockchain to match the scalability of information transfer today. Another way to get a rough estimation is this: there are 53 Terabytes of data transferred per second on the Internet. If there is one value transfer event associated with every 1KB information transferred(think about pay-per-byte), the TPS of an open financial system needs to reach 53 billion transactions per second.
Can blockchain scale to billions of TPS by just improving on-chain consensus algorithms?
On-chain consensus algorithm improvements aim to push the boundary of scalability-vs-decentralization tradeoff curves. It is an endeavor of picking the “sweet spot” on given tradeoff curves (e.g., dynamically picking the PoS stake lower limit within Ethereum Casper) or just use some different curves altogether (e.g., picking different consensus algorithms). Without going into potential improvement mechanisms, we can easily estimate the performance upper limit by just improving on-chain consensus algorithms. For any decentralized system, as long as every node needs to reach consensus on every single transaction, its performance has to be no better (in fact, much worse due to communication overhead) than a centralized system with a single node that processes every transaction, which is eventually bottlenecked by the processing capacity of the slowest node in the system. Even with sharding (arguably on-chain solution), the worst-case performance is also bottlenecked by such a bound with more convoluted tradeoffs we are not delving into this short blog.
The need for cSpeed off-chain operating network
But how does Internet reach its scalability? Well, because there is no large-scale consensus needed on the Internet. When your browser talks to Google, other billions of netizens do not need to validate this transaction. This simple insight leads to the question that whether we can bring most of the transactions off-chain, avoid on-chain consensus as much as possible and only resort to it in the worst case. To match the scale of Internet information transfer and support billions of transactions on blockchain with no compromise on trust, privacy, finality, or decentralization, we have to look beyond on-chain consensus algorithm improvements, and carefully create higher-level off-chain abstractions that gain scalability and come up with crypto-economic mechanisms to balance any tradeoffs. The involved technical tools can be generalized state channels, special or general sidechains, off-chain computing oracles and more. Though with promising potentials, these technologies are still in their infancy. Many fundamental challenges on both technology and economic mechanism design remain unsolved.
We believe the scope of such abstraction is often under-appreciated by the term of “Layer-2 solutions”. They are not band-aids or temporary solutions but are integral to the entire blockchain stack. This higher-level off-chain abstraction is essential for users to build and access scalable blockchain applications.
Therefore, we make the case for this abstraction to have a new name of Off-chain Operating Networks. Off-chain Operating Network can contain sophisticated technical layers and modules within itself and exhibits new kind of incentive and economic structure with various tradeoffs. To these points, we develop Celer Network, an off-chain operating network that fulfills the vision of cSpeed state transition and value transfer with our coherent technical and economic architecture.
In order to get the update from Celer Network’s series of blogs, please follow us on: