Scalability

Umbrella innovates on existing scaling solutions

There are three sides to scalability, Vertical Scaling, Horizontal Scaling, and Throughput Scaling.

Throughput Scaling

Commonly referred to as “the bottleneck” of a chain it’s the throughput of a single chain without any rollups or side chains. Umbrella can push the EVM to its limits with our DAG architecture, but will hit barriers that can be overcome by implementing new VMs or rebuilding Umbrella’s EVM in a more efficient programming language, for example, an EVM implementation built using Move.

Horizontal and Vertical Scaling

Using ZkRollups, Side Chains, and Realistic Rollups we can execute transactions off of the main chain to decrease costs and increase throughput 20-30x. This extreme speed will of course come at the expense of greater centralization which isn’t ideal for our ethos, but acceptable for certain applications. The greatest power of this scalability model is for devs to run their entire application on a dedicated sidechain. This makes the most sense for government applications such as CBDCs. Including these off-chain scaling methods, we should be able to take our base TPS to 1 million+. This will have variables that affect it, such as the size of each transaction’s data payloads, and would like to say again is theoretical and experimental. The best part is that, if someone creates a better scaling solution for EVM at any point we can implement it as an alternative and continue to build from there without disrupting the chain.

Proof of Relativity (PoR) - Predictive Transactions

Proof of Relativity is a mechanism for creating an on-chain score for addresses that can be used to mitigate issues with fraud in trustless environments. Having history, conducting transactions, locking funds, and adding value to the chain are all aspects that derive your score using AI/ML algorithms deployed on-chain. This score is kept private from others using Zero-Knowledge Proofs. This score is leveraged by nodes to create predictive finality for transactions and for application providers to offer advanced features, such as larger loan sizes or preferential liquidity

Realistic Rollups

Realistic Rollups is a novel concept for a scaling solution that executes transactions off the main Umbrella chain but uses the on-chain Relativity Score to preserve security before submitting batches back to Umbrella. It can be implemented to scale the rate at which transactions are processed while lowering the bulk cost. So developers with high-traffic applications can bundle transactions together to save on fees and processing time.

How it works:

The majority of the nodes verify transactions optimistically, while the minority of nodes verify transactions pessimistically. The exact number is based on the Relativity Scores of participants in each Epoch. A higher number of low-score participants will slightly increase the number of pessimistic nodes. Nodes rotate their role every “round” — an epoch. Optimistic nodes turn pessimistic and pessimistic nodes become optimistic. Transactions are split into various nodes based on their Relativity Score derived from the Proof of Relativity algorithm. After a batch of transactions is finalized, the nodes call Umbrella and submit a zero-knowledge rollup of the transaction.

Realistic Rollups break the existing Scaling Solutions landscape by being fraud and validity-proof while maximizing on-chain data.

Realistic Rollups would break this graphic by being both fraud and validity-proof while maximizing on-chain data.

Having a good score doesn’t mean your transaction skips validation. The validation isn’t as intense as if you’re a high-risk user (optimistic rollup). All transactions are still validated, and typical preventions against double-spend attacks, DDoS attacks, etc. are still in place. The goal here is to create predictability on the result of the transaction so the relativity score can help push the tx into finalization faster, not cancel the validation process.

Deciding Between Pessimistic and Optimistic using PoR

For example, transacting between Address A and Address B, where A & B both have a high relativity score will lead to it being verified optimistically. This can be done because there’s a high likelihood that these addresses have interacted with exchanges or known entities where they have been KYC'd. Therefore, there is a low likelihood of fraud. However, sending from Address A to Address B where A & B both have a low relativity score is verified pessimistically. This is done because there’s a high likelihood that these addresses are brand new and have never conducted any activity. Therefore, the actors could be attempting to hide their identity, increasing the likelihood of fraud.

Type of Optimistic Rollup

The Optimistic Rollup used within Realistic Rollups follows the standard set by Arbitrum, who we view as the leader in the EVM optimistic rollup space, and so, plan to contribute to their thought leadership in the matter by publishing and interacting with their opensource blockchain and community.

RAIN (Realistic Artificial Intelligence Network)

By combining Realistic Roll-Ups with our Enterprise-driven HDAG node architecture and our predictive Proof of Relativity mechanism, we can use RAIN as a Layer 2 scaling mechanism expected to 20-30x our Layer 1 speed - 120k tps to 2M-3M tps - this is highly experimental and theoretical and like most high scaling solutions introduces centralized aspects that rely on Umbrella's enable this speed.

What Our Implementation Looks Like:

Umbrella maintains the integrity of nodes by designing a Node system called a Hierarchical Directed Acyclic Graph (HDAG). More on that can be found below.

Out of all Supernodes on Umbrella, there is a subset of Supernova nodes; this is a slowly expanding number and will begin with the Umbrella team as the first maintainer. Supernova nodes + 1/10th of the number of Supernodes (infinite set) are tasked with running the Validator network. The 1/10th of supernodes rotate after several epochs have passed.

Realistic Rollups (explained above) verify transactions within a RAIN sidechain. There can theoretically be an infinite number of RAIN side chains operating in parallel to Umbrella. These chains can be set up as private or public networks and can alter the validator model as they see fit. They batch their data back to Umbrella and Umbrella increases that redundancy using our security design.

Relay Networks for Offline Transactions

Relay networks are very similar to Plasma because data is packed efficiently using Merkle trees. A Merkle root for a set of consecutive block headers is relayed, instead of each transaction. However, they differ in one particular aspect. Trustless relay networks replace the centralized relayer of Plasma with a decentralized relayer powered by a network of token stakers. Relay networks thus rely on creating proper incentives for token holders to stake their tokens, such that they can be trusted to run a relayer Litenodes act as the relay network for Umbrella. Litenodes speed up network transactions by relaying communications between two devices that are too far apart geographically to communicate directly. Litenodes share a portion of the network fee generated through the Relay Network. Litenodes on Umbrella utilize Phonon-enabled hardware to facilitate point-to-point communication between hardware without the need for a network connection.