<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><div><div>This is an announcement of <span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);">Faradawn Yang</span>'s MS Presentation</div><div><br></div><div>===============================================</div><div>Candidate: Faradawn Yang</div><div><br></div><div>Date: Wednesday, May 1, 2024</div><div><br></div><div>Time: 12 pm CT</div><div><br></div><div>Location: JCL 346</div><div><br></div><div>Title: <span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);">Leveraging AI for Faster Storage Access: a Graph-Neural-Network-Based Prefetcher</span></div><div><br></div><div>Abstract: <span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);">Despite the widespread adoption of SSDs in cloud environments, I/O access (disk reads and writes) remains a significant performance bottleneck. To mitigate access latency, various prefetching techniques have been developed. However, as servers increasingly host multi-user workloads, the efficacy of traditional prefetchers diminishes. Fortunately, with the ascendance of artificial intelligence, there is an escalating interest in utilizing machine learning methods to predict access patterns more accurately. Nonetheless, these methods face challenges: large model sizes, high inference latency, and unstable hit rates. In this paper, we introduce “Spectral Prefetcher”, a spectral graph neural network for rapid and precise predictions. This approach 1) reduces training and inference latency by 79.9%, 2) decreases memory usage by 33.3%, and 3) achieves a state-of-the-art hit rate, surpassing the baseline by 21.8%.</span></div><div><span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);"><br></span></div><div><br></div><div>Advisors: <span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);">Haryadi Gunawi</span></div><div><br></div><div>Committee Members: <span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);">Haryadi Gunawi, Suhail Rehman, and Cesar Stuardo</span></div></div><div><br></div><div>Draft:</div><div><br></div><div></div></body></html>