Sui is a blockchain network with a unique architecture and design. Its approach to handling assets and data structures sets it apart from other blockchain networks.
Asset-Oriented Architecture
Sui's execution environment is asset-oriented. This means that assets on Sui are more akin to programming resources and not data. Unlike data, which can be freely created, destroyed, copied, and moved, resources on Sui, once created, can only be moved from owner to owner, never copied or unintentionally destroyed or lost.
Object-Oriented Data Structure
Sui calls its data structure object-oriented. Rather than accounts, data is stored in Sui objects. These objects can be various entities such as decentralized exchanges (DEXs), auction contracts, or assets. Objects are packages, Sui’s term for smart contracts.
Transactions on Sui take live objects as input, close them off to future transactions, and produce mutated or newly created live objects as output. Sui’s global state is simply a collection of all live object states.
Transaction Parallelization
Sui requires transactions to specify dependencies. This is key for the network to determine a partial ordering for how to parallelize transactions. If separate transactions touch disjoint objects, they can skip ordering in the protocol and be processed in parallel, given their outcomes do not affect each other (that is, the actions are not dependent on each other).
Planned Storage Capabilities
Sui plans to add storage capabilities of arbitrary size. This feature will allow the network to store more data and handle more complex applications.
Validation and Scaling Throughput
Sui scales throughput differently than some other blockchain networks. It does not specify a finite amount of live objects that can be created, nor does it appear to localize validation. From what is understood, it appears transactions must interact with 2f+1 of the global validator pool by stake weight. However, Sui still scales horizontally, by INTRA-validator sharding.
While throughput capacity can decrease as new validators are added to the network (because transactions must interact with more validators), it can increase again if the added validator adds more compute resources.