The rapid expansion of Artificial Intelligence is fueling the growth of hyperscale data centers, which introduces significant challenges to existing power systems. This paper aims to provide a comprehensive survey of these integration challenges, specifically from the perspective of power grid utility. We find that AI data centers function as a distinct load category, characterized by high power density, rapid and large-scale power transients, and specific power quality profiles. These attributes create difficulties for long-term resource adequacy and transmission planning due to mismatched development timelines. They also strain real-time grid balancing and introduce risks to system stability, such as voltage and frequency deviations and converter-driven instabilities. The analysis further covers the economic and environmental footprints associated with this new type of consumer. The paper concludes that safely integrating these loads requires a coordinated strategy, encompassing data center-side technologies, grid-enhancing solutions, and new policy frameworks.
We present RQFTL, a demand-based FTL for mobile storage controllers that boosts the effective Logical-To-Physical (L2P) address translation cache capacity over state-of-the-art techniques. RQFTL stores a large part of the L2P cache in a compressed form, and employs a learned data structure called RQRMI that leverages tiny neural nets to quickly find the correct translation entry in the cache. RQFTL uses neural network inference for cache lookups, and rapidly retrains the neural nets to efficiently handle L2P cache updates. It is specifically optimized to achieve high coverage for scattered read accesses, making it suitable for popular read-skewed workloads such as mobile gaming. We evaluate RQFTL on hours-long real-world I/O traces of popular modern mobile apps, including games, video editing, and social networking apps collected on Google Pixel 6a phone. We show that RQFTL outperforms all the state-of-the-art FTLs in these workloads, increasing the effective L2P cache capacity by over an order of magnitude compared to DFTL and up to 5x over the recent LeaFTL. As a result, it achieves 65%, and 25% lower miss rate compared to DFTL and LeaFTL respectively, under the same SRAM capacity, and allows reduction of the total SRAM capacity of a controller by about a third of that of LeaFTL.
Practical experience is an important aspect of the training of every engineer. One way to develop such experience is by hands-on laboratory experimentation which involves cutting-edge technology. In this paper, we present a signal processing lab experiment developed for undergraduate students. Throughout the experiment, students first learn the basics of adaptive filtering and implement an adaptive echo cancellation algorithm using MATLAB. Then, students use a software framework for implementing the same echo cancellation algorithm on a BeagleBoard development platform using both ARM and fixed-point DSP cores of an embedded system-on-chip from Texas Instruments. In a relatively short amount of time (8 hours at the lab and a few additional hours of homework), students learn an important signal processing technique, as well as use complex state-of-the-art DSP hardware to implement it as part of an application running in real-time.