Dynamic Scoped Hierarchical Raft Accelerates Blockchain Consensus Across Heterogeneous Satellite Networks

Integrating blockchain technology into space-air-ground networks offers a resilient framework for decentralized data management, yet conventional consensus protocols struggle when deployed across heterogeneous orbital platforms subject to intermittent connectivity and asymmetric processing capacity. To resolve these operational constraints, researchers have developed Dynamic Scoped Hierarchical Raft, known as DSH-Raft, a consensus framework designed to accommodate performance disparities and unpredictable communication links inherent to satellite systems. By categorizing nodes according to measured performance and establishing tiered consensus subgroups, the protocol eliminates communication bottlenecks and service interruptions that plague traditional leader-based consensus in orbit. Experimental evaluations demonstrate that this scoped hierarchical architecture achieves substantial operational enhancements over standard implementations, boosting average transaction throughput by 65 percent, reducing confirmation latency by 12 percent, and cutting leader re-election delays by 71 percent during simulated network disruptions.

Distributed ledgers in space-air-ground communications reduce reliance on centralized ground stations, eliminating single points of failure while ensuring transparency, data immutability, and fault tolerance for collaborative tasks. However, classical consensus mechanisms exhibit severe vulnerabilities under strict hardware limits, volatile wireless links, and prolonged propagation delays of satellite orbits. Proof of Work protocols consume prohibitive electrical power and computational resources while displaying sluggish confirmation speeds, rendering them unsuitable for resource-constrained satellite nodes. Proof of Stake architectures reduce energy demands but risk concentrating governance control among wealthy token holders, while Practical Byzantine Fault Tolerance imposes a quadratic communication complexity of O(n squared) that severely restricts scalability across expanding networks. Although standard Raft offers simplicity, deterministic behavior, and lower message overhead, its uniform node handling and unconstrained leader elections create critical performance bottlenecks when applied to heterogeneous satellite infrastructures.

In standard Raft, an elected leader receives client transactions, sequences them into blocks, and replicates log entries across all follower nodes. In satellite constellations characterized by sharp disparities in computing power, bandwidth, and link stability, unconstrained elections often select underperforming nodes as cluster leaders. Such leaders become severe processing bottlenecks under high transaction volumes, while recurring orbital handovers trigger link disconnections that repeatedly force the system into election modes where transaction confirmation halts entirely. Requiring consensus from a majority of all nodes across the global cluster creates severe vulnerability during satellite transit scenarios, where simultaneous node disconnections prevent quorum formation. Previous Raft enhancements—including CRAFT erasure coding, OLR-Raft cached log addressing, cryptographic signature mechanisms, credit evaluation schemes, and multi-layered PBFT hybrids—either introduce excessive computational overhead or fail to resolve network partitioning and election delays across high-latency, heterogeneous environments.

To address these operational barriers, the researchers architected a multi-tiered blockchain environment integrating geostationary orbit satellites, low Earth orbit constellations, unmanned aerial vehicles, and terrestrial cloud servers. Geostationary satellites provide continuous broad-area coverage to maintain baseline network connectivity, while low Earth orbit satellites deliver primary high-speed data transmission services across orbital paths. Terrestrial cloud servers connect through stable ground networks to handle intensive computational loads, and unmanned aerial vehicles interface across orbital tiers to facilitate real-time coordination for environmental monitoring, logistics distribution, and disaster rescue operations. By combining wide satellite coverage with blockchain immutability and automated smart contracts, this space-air-ground network provides secure distributed collaboration among heterogeneous nodes. Operating within this integrated architecture, DSH-Raft structures consensus operations into three sequential phases comprising node performance evaluation, dynamic consensus group formation, and hierarchical transaction consensus.

In the performance evaluation phase, the system assesses each node by monitoring five specific network and computational metrics. The evaluation model treats network latency as a negative indicator while designating node bandwidth, satellite window time, single-core processor score, and available processor cores as positive indicators. Each raw metric undergoes min-max normalization to project values onto a uniform scale from zero to one before being combined into an overall score through a weighted linear sum. Although the model permits dynamic weight adjustment—such as elevating latency weighting in environments with frequent satellite link handovers—the experimental validation assigned equal unitary weights across all five parameters. A distributed monitoring service deployed across the network continuously recalculates these composite performance scores and periodically publishes a global ranking table, allowing all nodes to independently verify their current standing relative to the dynamic consensus threshold.

During consensus group formation, nodes are partitioned into M consensus nodes and N minus M non-consensus nodes according to their evaluated scores. To balance operational workloads between the primary leader and subgroup master nodes, M is set to the ceiling of the square root of total nodes N, balancing subgroup count against individual subgroup membership. In configurations of four or fewer nodes, the algorithm naturally degenerates into classical Raft. In larger networks, the top-ranked consensus node is elected as the primary leader, while the remaining M minus one consensus nodes serve as master nodes for discrete consensus subgroups. Non-consensus nodes are sequentially assigned across these M minus one subgroups through a round-robin grouping algorithm operating with linear time and space complexity O(N). This hierarchical division restricts leader candidacy exclusively to top-performing consensus nodes while insulating the leader from direct communication with the entire network.

The consensus phase executes transactional processing through this hierarchical pipeline, confining voting and commitment operations strictly to consensus nodes. Client transactions submitted to any node are forwarded directly to the leader, which encapsulates each transaction into a Raft log entry marked with the current term and log index. The leader replicates this entry exclusively to consensus nodes, which verify that the term and index align with their local log state. A log entry is committed once the leader receives confirmations from a majority quorum, defined as floor(M/2) + 1 consensus nodes. Upon reaching quorum commitment, the leader aggregates the committed entries into a block and broadcasts it to all consensus nodes. Each consensus node validates the block, writes it to its local ledger, and propagates it downward to the non-consensus nodes within its assigned subgroup, ensuring synchronized network-wide data consistency without overloading the central leader.

To guarantee fault tolerance across volatile orbital links, DSH-Raft defines targeted recovery protocols for three distinct failure scenarios. When the leader node disconnects, consensus nodes detect the failure upon heartbeat timeout, notify their non-consensus subgroups, transition to candidate status, and request votes from the latest top-M ranked nodes. Candidates securing at least floor(M/2) + 1 votes win the election, while non-consensus nodes newly elevated into top-M ranking via updated performance tables can likewise solicit votes. If a non-consensus node disconnects, the core consensus process remains unaffected, and the node recovers missing logs from its subgroup master upon reconnecting. If an intermediate consensus subgroup master fails, the leader detects the unacknowledged heartbeats and promotes the highest-ranked non-consensus node within that subgroup to master, downgrading the failed node and preserving uninterrupted block distribution.

Theoretical analysis verifies that DSH-Raft preserves all core safety properties of classical Raft, encompassing election safety, leader completeness, log matching, and state machine safety. Election safety is maintained because each consensus node grants only one vote per term, ensuring that overlapping majority quorums within consensus set C prevent dual leadership. Leader completeness is guaranteed by quorum intersection, which ensures any newly elected leader contains all entries committed in prior terms. Log matching is enforced through strict prefix consistency checks during log replication, while state machine safety is preserved because commitment decisions depend solely on consensus quorums, allowing non-consensus nodes to safely install verified committed state. When partitions isolate a leader within a minority of consensus nodes, commitment halts, preserving safety until reconnection. Linearizable reads are supported through leader or quorum queries, while joint consensus mechanisms safeguard cluster transitions during node promotions and demotions.

Researchers evaluated DSH-Raft on an experimental platform comprising three cloud servers and seventeen virtual machines running a blockchain framework implemented in Golang. Network topologies ranging from 4 to 20 nodes were tested to assess throughput, latency, and recovery under varying scales. Throughput and transaction latency benchmarks configured three consensus groups, featuring a primary consensus committee of four nodes including the leader. Node performance heterogeneity was modeled around a mean score of 0.5 using the coefficient of variation (CV), evaluated across values from 0.05 for near-uniform environments to 0.5 for pronounced node discrepancies. Leader election resilience was tested during one hour of continuous execution under simulated large-scale node disconnections, benchmarking DSH-Raft against classical Raft, OLR-Raft, and CRaft under identical network conditions.

The experimental results demonstrate that DSH-Raft achieves substantial performance gains over alternative consensus algorithms as network size expands. In a twenty-node cluster, DSH-Raft improved average transaction throughput by 65 percent and reduced transaction latency by 12 percent compared to classical Raft, consistently outperforming OLR-Raft and CRaft. At a network scale of four nodes, DSH-Raft exhibited identical latency and throughput to classical Raft because consensus grouping does not trigger below five nodes, naturally causing the system to execute standard Raft procedures. Under simulated large-scale node disconnections with high performance variance (CV = 0.5), DSH-Raft reduced average leader election time by 71 percent relative to standard Raft. This acceleration occurs because consensus node classification considers historical stability and satellite window time, reducing leader failure probabilities and ensuring sufficient operational consensus nodes remain available to complete rapid re-elections.

Subject of Research: Hierarchical Raft consensus optimization for blockchain systems in heterogeneous satellite networks

Article Title: Satellite network-optimized Dynamic Scoped Hierarchical Raft for blockchain consensus

Article References: Li, J., Huang, R., Shi, N., Weng, H., Wang, Q., Li, G., He, Y., & Tian, C. (2026). Satellite network-optimized Dynamic Scoped Hierarchical Raft for blockchain consensus. Autonomous Intelligent Systems, 6(1), Article 4. https://doi.org/10.1007/s43684-026-00126-3

Image Credits: AI Generated

DOI: 10.1007/s43684-026-00126-3

Keywords: blockchain consensus, satellite networks, Raft algorithm, distributed ledgers, hierarchical consensus, network latency, fault tolerance, throughput optimization, Satellite, network-optimized, Dynamic, Scoped