Vladimir Korkhov

Data transfer between organizations over untrusted public networks is associated with risks of data loss, corruption, and unauthorized access. To address this problem, we present a message broker prototype based on Hyperledger Fabric using a two-layer architecture: PBFT partitions for data transfer and Raft topics for system management. Only metadata is stored in the blockchain, while messages are placed in external storages. This approach provides higher fault tolerance and scalability compared to existing traditional and blockchain-based solutions.

The subject of this research is ensuring data consistency and fault tolerance in distributed systems using consensus algorithms. The object of the study is the Paxos consensus algorithm, which is widely used to coordinate decisions between nodes in distributed environments. Despite its reliability, Paxos has several practical limitations that may affect system performance and availability under unstable network conditions. The paper analyzes several shortcomings of the classical Paxos algorithm, particularly the difficulty of gathering a quorum when some nodes fail or become temporarily unavailable, as well as performance degradation caused by the need to involve all acceptors in the consensus process. Special attention is paid to the influence of message loss and network instability on the efficiency of consensus algorithms in distributed systems. To address these challenges, the study proposes a modified version of the Paxos algorithm called Orchestra-Paxos. The proposed modification focuses on achieving partial consistency while improving system fault tolerance under conditions of failures and unstable network communication. In the proposed approach, the participation of more than half of the accepting nodes is not strictly required in each replication round. Consensus can be reached based on responses from an active subset of nodes, while other nodes are temporarily marked as unavailable. The research methodology is based on a comparative analysis of Orchestra-Paxos and Multi-Paxos using distributed system modeling and experimental performance evaluation. The experiments measure the average replication round time, the total completion time, and the number of rounds required to accept all proposals. The results show that Orchestra-Paxos significantly improves performance under high message loss conditions, reducing the required number of rounds and the overall completion time by an average factor of 12 compared to Multi-Paxos when the probability of message loss reaches 60%.

Effective road crack detection is crucial for road safety, infrastructure preservation, and extending road lifespan, offering significant economic benefits. However, existing methods struggle with varied target scales, complex backgrounds, and low adaptability to different environments. This paper presents the BsS-YOLO model, which optimizes multi-scale feature fusion through an enhanced Path Aggregation Network (PAN) and Bidirectional Feature Pyramid Network (BiFPN). The incorporation of weighted feature fusion improves feature representation, boosting detection accuracy and robustness. Furthermore, a Simple and Effective Attention Mechanism (SimAM) within the backbone enhances precision via spatial and channel-wise attention. The detection layer integrates a Shuffle Attention mechanism, which rearranges and mixes features across channels, refining key representations and further improving accuracy. Experimental results show that BsS-YOLO achieves a 2.8% increase in mean average precision (mAP) for road crack detection, supporting its applicability in diverse scenarios, including urban road maintenance and highway inspections.