METHOD FOR LEADER NODE SELECTION AND PROCESSING PIPELINE FORMATION IN DISTRIBUTED TELECOMMUNICATION SYSTEMS

The paper proposes a method for selecting the master node (coordinator) in distributed telecommunication systems (DTS) with a clustered architecture and pipeline-based data processing. The method is aimed at ensuring stable data flow management within a cluster under conditions of dynamically changing workloads, network instability, and limited computational resources. Unlike classical leader election procedures that rely on global synchronization or broadcast-based voting algorithms, the proposed approach implements deterministic coordinator selection based on local node ranking, taking into account latency metrics, computational capacity, and unique identifiers. As part of the proposed method, the Gossip-based leader election algorithm has been enhanced by integrating mechanisms for periodic metric exchange, local candidate ranking, pre-assignment of backup nodes, and automatic control transfer in case of coordinator failure. The algorithm maintains the current state of nodes as local lists and uses heartbeat-type control messages to confirm the coordinator’s activity. To prevent redundant message propagation, mechanisms such as Time-To-Live (TTL) and iteration markers have been introduced, which eliminate the circulation of outdated data. Experimental modeling shows that the improved algorithm achieves full data convergence in a 50-node cluster within 4–6 seconds, demonstrates high resilience to message loss (≤1%), and ensures minimal delay during automatic coordinator reassignment. Compared to the fast Bully algorithm, the proposed approach reduces total control recovery traffic by up to 18% and significantly improves cluster stability under frequent topology changes. Thus, the proposed method enables effective management of clustered DTS with pipeline processing without initiating explicit election procedures, making it suitable for deployment in scalable and high-load telecommunication environments.