Authors: Srujana Parepalli
Abstract: Performance optimization of enterprise data pipelines has traditionally relied on rule-based heuristics, manual tuning cycles, and the accumulated intuition of experienced practitioners; however, as data volumes scale into terabytes and petabytes, workloads become increasingly heterogeneous, and execution environments span databases, ETL engines, and distributed infrastructure, these approaches struggle to deliver consistent and timely results. This paper presents an AI-assisted performance tuning framework for PL/SQL execution environments and Informatica PowerCenter workflows that augments established database performance metrics such as execution plans, wait events, resource utilization, and ETL session statistics with machine-learning-driven optimization techniques capable of learning from historical workload behavior. Building on foundational research in automatic database tuning, self-managing and autonomic systems, and ETL performance engineering, the proposed architecture continuously correlates workload characteristics, configuration parameters, and observed performance outcomes to generate data-driven recommendations for optimal SQL execution strategies, memory and session configurations, partitioning schemes, and workflow design patterns. By synthesizing academic research and industry practices published between 2000 and 2017, the study illustrates how AI-based optimization complements traditional tuning methods by reducing manual intervention, improving adaptability to changing data patterns, and delivering measurable improvements in throughput, latency, and operational stability across large-scale enterprise data platforms.
