To address the problem of expired memories still participating in current retrievals after explicit state updates in long-running agents, while historical versions need to be continuously retained, this study proposes a version validity-driven persistent memory dual-storage mechanism, VPM-DS. This method maintains the current/historical state using an entity-attribute version chain. Through constraints of current state uniqueness, historical integrity, and temporal monotonicity, memory validity is mapped to retrieval participation eligibility, and a relationship model between the historical working set and retrieval latency is established. This study validates the mechanism based on Memory Remembering, LongMemEval-Cleaned Knowledge Update, and continuous write loads. Under quarterly conditions, VPM-DS only improves FAMA by 0.3 percentage points compared to State-aware Single-Store; under 9k conditions and a 30% historical version ratio, P95 decreases by 10.8%. The results show that version identification contributes the main semantic gains, and the independent value of dual storage is mainly reflected in the control of the active working set under high historical loads.

Shefeng Li. Persistent Memory Mechanisms and Storage Architecture Design for Agent-based Information Systems. International Journal of Big Data Intelligent Technology (2026), Vol. 7, Issue 3: 31-41. https://doi.org/10.38007/IJBDIT.2026.070304.

[1] Zhang Z, Dai Q, Bo X, et al. A Survey on the Memory Mechanism of Large Language Model-based Agents[J]. ACM Transactions on Information Systems, 2025, 43(6): Article 155, 1-47. DOI: 10.1145/3748302.

[2] Zheng S, He K, Yang L, Xiong J. MemoryRepository for AI NPC[J]. IEEE Access, 2024, 12: 62581-62596. DOI: 10.1109/ACCESS.2024.3393485.

[3] Wang K, Yan S, Yuan H, et al. Toward Interpretable and Persistent Personalization: A Memory-Augmented Agent Framework for LLM-Based Travel Planning[J]. IEEE Access, 2025, 13: 193125-193141. DOI: 10.1109/ACCESS.2025.3630495.

[4] Wang, Z. (2026). Research on Operation Data Analysis and Efficiency Improvement of the New Energy Industry Based on Machine Learning. Procedia Computer Science, 285, 605-612.

[5] Cui, N. (2024, December). Research on Adaptive Urban Traffic Green Wave Control Algorithm and Safety Based on Reinforcement Learning DQN-COP. In International Conference on Big Data Analytics for Cyber-Physical System in Smart City (pp. 379-387). Singapore: Springer Nature Singapore.

[6] Wang, B. (2025, November). Research on Performance Improvement of Personalized Recommendation System for E-commerce Platform Based on Big Data and Collaborative Filtering Algorithm. In International Conference on Computational Technologies for Research in Data Analytics (pp. 36-46). Cham: Springer Nature Switzerland.

[7] Hou, Y. (2026, March). Adaptive Upgrading of Computing Power Platforms for Large Model Workloads. In International Conference on Computing and Communication Systems for Industrial Applications (pp. 429-437). Cham: Springer Nature Switzerland.

[8] Xin, H. (2026). Research on Distributed Data Pipelines and Event-Driven Architecture for Freight Visibility.

[9] Zhang, Z. (2026). User Behavior Anomaly Identification Integrating Telemetry Data and Its Application in Consumer Finance Risk Control.

[10] Jin, C. (2026). Application of Data Mining Technology in Predicting User Churn and Retention Strategies in E-Commerce.

[11] Liu, S. (2026). Design and Implementation of Retrieval Enhancement Generation Method for Enterprise Knowledge Base.

[12] Zhang, Z. (2026). Research on Service-oriented Packaging of AI Systems and Data-driven Decision Support for Real-world Business Processes. Engineering Advances, 6(3).

[13] Zhang, Y. (2025, January). Research on the Application and Optimization of Multi Dimensional Data Model Based on Kylin in Enterprise Technology Management. In International Conference on Innovative Computing (pp. 234-241). Singapore: Springer Nature Singapore.

[14] Qian, X. (2026, July). Application of the Combination Model of Prophet and LSTM in Multi-Level Inventory Demand Prediction for Retail Supply Chain of CPG-Consumer Packaged Foods. In 2026 5th International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE) (pp. 1-7). IEEE.

[15] Jin, C. (2026). Construction and Empirical Study of Credit Risk Assessment Index System for Small and Micro Enterprises in Commercial Banks.

[16] Liu, S. (2026). Automatic Generation Method and Implementation of Web Front-End Components Based on Generative AI.

[17] Zhang, Z. (2026). User Behavior Device Reliability Joint Modeling and Intelligent Risk Control Practice In Consumer Technology.

[18] Zhang, Y. (2026). Exploration of Enterprise Big Data Microservice Architecture Based on Domain-Driven Design (DDD). Procedia Computer Science, 282, 1994-2003.

[19] Jingzhi Yin. (2026). Constructing a CMBS Default Risk Sentiment Index Using Financial Text Embeddings and Evaluating Its Predictive Effectiveness. Engineering Advances, 6(1), 50-54.

[20] Fei, W., Xue, W., & Wang, B. (2025). Research on the innovative path of rural education teaching methods from the perspective of PBE and PBL integration. Int. Acad. J. Humanit. Soc. Sci.

[21] Fei, W. (2026). Families Of Chinese Background In Rural Alabama: Education, Immigration, And Identity.

[22] Wendan Xue, Wen Fei. An Analytical Study on the Differences and Determinants of Basic Education Resource Allocation in Rural China and the United States. International Journal of Educational Research and Development, 1(3), 13-17.

[23] Boli Wang, Wen Fei. Innovations and practices in the classroom for rural aesthetic education in the perspective of school integration. International Journal of Educational Research and Development, 3(1), 36-41.

[24] Fei, W., & Qian, J. (2026). Application and Exploration of Reverse Instructional Design in the Teaching of the Primary School Physical Education" Walking and Playing" Unit. Journal of Sociology and Education, 2(2).