Traditional blockchain architectures, such as Bitcoin, enforce a strict, linear sequence of blocks. In these systems, a single block producer extends the chain at any given time, and concurrent block creation results in "forks" that are treated as conflicts or wasted computational effort. As analyzed by Aviv Zohar in Securing and scaling cryptocurrencies, this linear constraint creates a fundamental trade-off between security and performance: increasing transaction throughput or reducing block times exacerbates network delays, leading to accidental forks that compromise the system's integrity.
To bypass this sequential bottleneck, researchers have turned to Directed Acyclic Graph (DAG) ledger topologies. Instead of organizing transactions into a single-file line, a DAG allows the ledger to grow horizontally. In this paradigm, concurrent block production is treated as a standard state rather than an anomaly, enabling parallel transaction processing and asynchronous consensus.
The Intuition of Multi-Parent Topologies
The core structural shift from a blockchain to a DAG lies in how blocks reference historical data. In a traditional chain, each block contains a single pointer to its immediate parent. In a DAG ledger, a new block or transaction references multiple preceding entries simultaneously.
This multi-reference topology fundamentally alters how the network scales. As Bin Cao 0002 et al. point out in When Internet of Things Meets Blockchain: Challenges in Distributed Consensus, allowing multiple ledger branches to grow and merge simultaneously enables horizontal scaling. Instead of waiting for a fixed block generation interval, multiple nodes can validate and append transactions concurrently. Consequently, the network's processing capacity scales proportionally with the number of active participants, a property highly desirable for high-frequency environments like the Internet of Things (IoT) or vehicular networks (Wenhui Yang et al., LDV: A Lightweight DAG-Based Blockchain for Vehicular Social Networks).
Furthermore, this structure changes the incentive and cost dynamics. In some DAG designs, such as the IOTA Tangle discussed by Sabah Suhail et al. in Orchestrating product provenance story: When IOTA ecosystem meets electronics supply chain space, a user must validate two previous transactions before issuing their own. This self-sustaining mechanism removes the distinction between users and dedicated miners, eliminating transaction fees and enabling high-volume micro-transactions.
The Ordering Challenge: Resolving the Non-Linear State
While DAGs excel at high-throughput data ingestion, their non-linear nature introduces a severe challenge: transaction finality and total ordering.
In a linear blockchain, the order of transactions is implicitly defined by the order of the blocks. In a DAG, because transactions are processed in parallel across disparate branches, determining a globally consistent sequence becomes significantly more complex. As Yang Xiao et al. explain in A Survey of Distributed Consensus Protocols for Blockchain Networks, without a single-producer bottleneck, the network requires sophisticated mechanisms to resolve conflicts (such as double-spends) and establish a deterministic state. Furthermore, non-linear structures are susceptible to unique failure modes, such as balancing attacks, where an adversary splits the network's hashing power across parallel branches to keep them perpetually tied, delaying finality.
Mechanisms for DAG Consensus
To achieve a stable, deterministic state without sacrificing parallelism, researchers have developed several distinct consensus strategies within DAG structures.
1. Pairwise Voting and Topological Sorting
One approach is to infer ledger consistency directly from the graph's geometry via voting mechanisms. For instance, the SPECTRE protocol, highlighted by Aviv Zohar, uses a pairwise voting mechanism derived from the causal relationships between blocks. Replicas can determine the final execution sequence by performing topological or reverse topological sorts on conflict graphs, allowing non-interfering operations to run in parallel while deterministically resolving conflicts (Michael J. Whittaker et al., SoK: A Generalized Multi-Leader State Machine Replication Tutorial).
2. Hybrid Linear Backbones
Another prominent strategy is to embed a traditional, linear consensus backbone within the highly parallel DAG. Jiahao He et al. propose this hybrid architecture in Consensus Mechanism Design based on Structured Directed Acyclic Graphs. In their design, blocks utilize a multi-reference topology containing three distinct pointers: one to maintain personal history, one to ensure graph connectivity, and a third that links to a high-difficulty consensus "milestone" chain. This architecture allows the network to enjoy high-throughput parallel block production for data ingestion, while relying on periodic, high-difficulty milestones to anchor a definitive linear ordering for security.
3. Decoupling Dissemination from Ordering
Modern DAG consensus protocols often achieve high performance by completely separating the process of spreading information from the process of establishing a final sequence. As noted by Huan Yu Wu et al. in When Distributed Consensus Meets Wireless Connected Autonomous Systems: A Review and a DAG-Based Approach, this decoupling allows nodes to build a local history of interactions asynchronously, which can be reconciled over time.
By integrating reliable broadcast (RBC) primitives directly into the DAG structure, nodes can broadcast vertices in rounds. As demonstrated by Nibesh Shrestha et al. in Optimistic, Signature-Free Reliable Broadcast and Its Applications, nodes can observe the propagation of broadcast messages within the graph to commit leader vertices with minimal additional network steps, achieving high-throughput consensus without relying on heavy cryptographic signatures.
Summary of the Paradigm Shift
| Dimension | Traditional Linear Blockchain | Directed Acyclic Graph (DAG) Ledger |
|---|---|---|
| Topology | Sequential chain of single-parent blocks | Non-linear graph of multi-parent blocks/transactions |
| Concurrency | Concurrent blocks cause forks (treated as errors) | Concurrent blocks are standard state (processed in parallel) |
| Scalability | Limited by block intervals and propagation delays | Scales horizontally with the number of active participants |
| Primary Challenge | Throughput bottlenecks and high latency | Complex transaction ordering and finality resolution |
By shifting the fundamental data structure from a line to a graph, DAG-based ledgers decouple data availability from strict ordering. This structural evolution allows distributed systems to bypass the sequential bottlenecks of early blockchain designs, offering a scalable foundation for high-concurrency applications.
Key Foundational Work
Securing and scaling cryptocurrencies (2017) — contributed 3 concept elements to this direction.