Memory Disaggregation for AI Infrastructure

This systems review examines memory disaggregation for artificial-intelligence infrastructure. The organizing question is when pooled and remote memory improves utilization without unacceptable latency, interference, or recovery complexity. Ten related scholarly sources are synthesized through a decision-centered framework spanning problem definition, mechanism, measurement, evaluation, implementation, and governance. The review does not invent experiments, pooled estimates, or unreported quantitative results. It instead evaluates the strength and transferability of the available evidence, with particular attention to comparing capacity gains without modeling workload locality and failure domains. The resulting framework links technical or empirical performance to explicit use conditions and identifies tests that should precede wider adoption in large-scale AI training and inference platforms.

1. Aguilera, M. K., Amaro, E., Amit, N., Hunhoff, E., Yelam, A., & Zellweger, G. (2023). Memory disaggregation: why now and what are the challenges. ACM SIGOPS Operating Systems Review, 57(1), 38-46. https://doi.org/10.1145/3606557.3606563 DOI
2. Heo, T., Kang, S., Lee, S., Hwang, S., & Huh, J. (2021). Hardware-assisted Trusted Memory Disaggregation for Secure Far Memory. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2108.11507 DOI
3. Kwon, M., Jang, J., Choi, H., Lee, S., & Jung, M. (2023). Failure Tolerant Training With Persistent Memory Disaggregation Over CXL. IEEE Micro, 43(2), 66-75. https://doi.org/10.1109/mm.2023.3237548 DOI
4. Maruf, H. A., & Chowdhury, M. (2023). Memory Disaggregation: Advances and Open Challenges. ACM SIGOPS Operating Systems Review, 57(1), 29-37. https://doi.org/10.1145/3606557.3606562 DOI
5. Mehra, P., & Coughlin, T. (2022). Taming Memory With Disaggregation. Computer, 55(9), 94-98. https://doi.org/10.1109/mc.2022.3187847 DOI
6. Quek, Y. T., Woo, W. L., & Logenthiran, T. (2019). Load Disaggregation Using One-Directional Convolutional Stacked Long Short-Term Memory Recurrent Neural Network. IEEE Systems Journal, 14(1), 1395-1404. https://doi.org/10.1109/jsyst.2019.2919668 DOI
7. Wang, R., Wang, J., Idreos, S., Özsu, M. T., & Aref, W. G. (2022). The case for distributed shared-memory databases with RDMA-enabled memory disaggregation. Proceedings of the VLDB Endowment, 16(1), 15-22. https://doi.org/10.14778/3561261.3561263 DOI
8. Wang, Z., Guo, Y., Lu, K., Wan, J., Wang, D., Yao, T., & Wu, H. (2024). Rcmp: Reconstructing RDMA-Based Memory Disaggregation via CXL. ACM Transactions on Architecture and Code Optimization, 21(1), 1-26. https://doi.org/10.1145/3634916 DOI
9. Xia, M., Liu, W. A., Wang, K., Song, W., Chen, C., & Li, Y. (2020). Non-intrusive load disaggregation based on composite deep long short-term memory network. Expert Systems with Applications, 160, 113669. https://doi.org/10.1016/j.eswa.2020.113669 DOI
10. Yu, X., Jia, X., Zhang, J., Wang, Y., Yu, S., & Qi, Z. (2025). Rethinking Virtual Machines Live Migration for Memory Disaggregation. IEEE Transactions on Parallel and Distributed Systems, 36(11), 2310-2324. https://doi.org/10.1109/tpds.2025.3597149 DOI