Existing research on Content Delivery Networks (CDNs) predominantly focuses on optimizing the delivery of hot content—popular items that attract frequent and repeated access from large user bases. However, at Meta, we have identified that cold content, which is less popular and accessed infrequently, poses significant challenges for user experience, especially with large volumes of direct messaging media. Through extensive analysis of Meta's CDN data, we quantify the impact of cold content performance on user experience and reveal substantial differences across geographic regions. Motivated by these findings, we propose CacheFlare, a suite of production-deployed solutions designed to enhance Quality of Experience (QoE) by increasing CDN hit rates for cold content. Through experiments and production validation, we show that even cold content can benefit from improved caching strategies, yielding up to a 10% improvement across a range of network performance metrics.

This paper studies low-complexity robust beamforming in mobile massive multiple-input multiple-output (MIMO) wireless communication systems, which introduces a robustness factor and deep learning (DL)-based framework to mitigate the impact of imperfect channel state information (CSI). By incorporating the estimation uncertainty, we aim to design transceivers to minimize outage probability subject to a total transmit power constraint. However, due to the probabilistic constraints, the optimization problem is difficult to solve. Therefore, we introduce a robustness factor to the quality of service (QoS) constraints by maximizing a zero-outage region, based on an extension of the offset maximization method. Then, we convert the problem into a convex problem and get a quasi-closed-form solution. Besides, iterating the fixed point equation of the Lagrange multipliers in the inequality optimization introduces enormous computational complexity. To this end, we present a novel DL-based algorithm to predict the multipliers directly from imperfect CSI, which includes a column convolutional layer elaborately designed for channel input. Comprehensive experimental comparisons demonstrate the proposed robust transceiver can improve the outage probability significantly over conventional baselines, and the DL-based algorithm can further reduce the running time.

Thermal issues have become critical roadblocks for the development of advanced chip-multiprocessors (CMPs). In this paper, we introduce a new angle to view transient thermal analysis - based on predicting thermal profile, instead of calculating it. We develop a systematic framework that can learn different thermal profiles of a CMP by using an autoregressive (AR) model. The proposed AR model can serve as a fast alternative for predicting the transient temperature of a CMP with reasonably good accuracy. Experimental results show that the proposed AR model can achieve approximately 113X speed-up over existing thermal profile estimation methods, while introducing an error of only 0.8°C on average.

Full-chip thermal monitoring is an important and challenging issue in today's microprocessor design. In this paper, we propose a new information-theoretic framework to quantitatively model the uncertainty of on-chip temperature variation by differential entropy. Based on this framework, an efficient optimization scheme is developed to find the optimal spatial locations for temperature sensors such that the full-chip thermal map can be accurately captured with a minimum number of on-chip sensors. In addition, several efficient numerical algorithms are proposed to minimize the computational cost of the proposed entropy calculation and optimization. As will be demonstrated by our experimental examples, the proposed entropy-based method achieves superior accuracy (1.4x error reduction) for full-chip thermal monitoring over prior art.