MATHEMATICAL MODEL AND ADAPTIVE CONTROL OF A DATA LOG MANAGEMENT ALGORITHM UNDER VARIABLE SERVER LOAD CONDITIONS

Abstract

The article addresses the problem of optimizing the logging of business process execution states and managing outdated logs in a relational database management system . A mathematical model and an algorithm for dynamic log management are proposed, based on an adaptive processing threshold and scheduling parameters. The primary advantage of the approach lies in the algorithm’s ability to adapt to server workload fluctuations by increasing the number of processed records during periods of low system activity and minimizing performance impact during peak loads. The algorithm adjusts the number of records to be deleted and the frequency of operations in response to changes in server load. The key efficiency criterion is the balance between the volume of processed data, operation execution time, and system performance impact. Experimental results demonstrate the algorithm’s effectiveness in achieving a balance between system performance and the amount of processed data. The experiments show that, due to self-adaptation, the algorithm increases the number of processed records during periods of low system activity, while reducing its impact on performance during peak loads. The proposed methodology is promising for applications in business process automation, particularly in CRM systems (Customer Relationship Management), ERP systems (Enterprise Resource Planning), and e-commerce platforms, where stable server performance is critical. The use of relational databases combined with the proposed algorithm reduces implementation and training costs while maintaining high system flexibility and scalability. Comparison with well-known approaches highlights the relevance and prospects of applying adaptive methods for managing background tasks (particularly log cleaning), driven by the need for the efficient utilization of computational resources and compliance with data retention and availability requirements. Future research directions include extending the algorithm to support multi-user systems and integrating it with predictive analytics tools.