DAG (directed acyclic graph) - is a structure which is used in Cellframe zero-chain. DAG consists of events, each event can contain only one data type (datum). Events are collected in sets, form connected graphs, and are called a Round. The round on forming stage is placed in the Global Database (GDB), not in the chains, then is carried from GDB to the chains after the minimum time has passed, if it becomes connected.
The DAG, instead of a clear sequence of blocks, writes information non-linearly, without a sequential chain of blocks. The reliability of the algorithm is mathematically described as Asynchronous Byzantine Error Tolerance (ABFT). Algorithms are very difficult to hack, and more than two-thirds of the participants need to agree to make a decision at the network level. Due to DAG, the hashgraph has all the advantages of the blockchain, while demonstrating a higher transaction performance. In general, the hashgraph works as follows: network participants (nodes) exchange data over the network, information about the transaction is sent to two random nodes, transmitting them, in turn, to two other nodes - and so on exponentially, until the volume of informed nodes becomes sufficient for verification of the transaction; nodes exchange with each other only transaction data, and not all information about the network; transactions are recorded in chronological order - you can track their history.
In the hashgraph, each node knows the entire transaction history, therefore, to achieve consensus, “virtual voting” is used - there is no need to coordinate all the nodes with each other: each of them already knows how the other node will “vote”. Other advantages of DAG are: scalability and performance - up to 500 thousand transactions per second (potentially - up to a million); an increase in the number of network participants increases the bandwidth; low resource consumption and high resistance to DDos attacks and botnets; no need for mining and Proof of Work (PoW); commissions close to zero.