Dynamic workloads running on multiple hosts will bring changing access patterns on CXL-enabled shared disaggregated memory. Existing works often un-traceably cache multi-source accesses, making it hard to exploit each host’s access behavior and assure service quality. Our solution Alchemy jointly optimizes cache replacement and bypassing and runs as an online reinforcement learning agent with source-aware adaptivity. It gives rewards derived from sampling-based action effectiveness and per-host macro performance. The multi-host prototype-based results on FPGAs show $8.71 \%-14.56 \%$ reduction in average access latency over LRU policy and 44x faster than the hardware-efficient ICGMM method in decision-making with comparable overhead.

Traditional logging methods often encounter complications like collapses of wellbores and instrument malfunctions, leading to the loss or misalignment of logging data. Re-logging is expensive, particularly for oil and gas wells that have already been cemented; therefore, reconstructing logging curves is a more economical option. Although traditional methods like empirical modeling and multiple fitting are effective in some cases, they are limited by geographic location and rock type, making it difficult to meet the demands for precise interpretation of logging data and detailed description of reservoirs. In this study, a CNN-BiLSTM-Attention model is proposed that combines a bidirectional long and short-term memory network (BiLSTM), a convolutional neural network (CNN), and an integrated attention mechanism. The model optimizes the ability of traditional deep learning methods to process time-series data and effectively establishes complex nonlinear mapping relationships between inputs and outputs. In this paper, root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R-squared) are used to evaluate the performance of the CNN-BiLSTM-Attention model. The results demonstrate that the divergence between the reconstructed value and the actual value is negligible in the log reconstruction experiment of CNN-BiLSTM-Attention, particularly under difficult geological settings. This study not only provides an innovative solution for the accurate prediction and effective reconstruction of logging curves but also provides valuable practical experience and a foundation for the implementation of deep learning in geophysics.