CS244c project presentations
Timing
Please prepare a 12-minute presentation and demo, which will leave 2-3 minutes for questions.
Venue and timing
Presentations will be in CoDa W415. All students are encouraged to attend all presentations. Lunch will be served.
Tuesday, March 17
9:00am
Investigating DCTCP Performance Beyond the Data Center. Alisha Nanda and Shashwati Shradha. code. paper.
Striped Tunnels: Reliable Transport from Unreliable Paths. Carlos Cecilio Hernandez-Meza, Aditya Sriram, Vidur Gupta, and Sebastian Beckett Fisher. code. paper.
Residential proxy detection. Sophia Barnes and Stephanie Vezich Tamayo. code. paper.
Generalization in Learned Congestion Control: Decision Trees, Deep RL, and LLM-Guided Evolution. Duy Nguyen, Andrew Grant, and Armaan Abraham. code. paper.
Approximating Central Rate Limits for Distributed Traffic. Matthew Torre, Amy Chang, and Andy Wang. code. paper.
Reproducing and Extending EyeNexus - Gaze-adaptive Encoding for VR Streaming. Sam Jett and Rohan Parekh. code. paper.
10:30am Break
10:45am
Reproducing Deep Fingerprinting: A Replication Study of Deep Learning-Based Website Fingerprinting Attacks. Devin Ty Fung, Miro Ohannes Swisher, Yousef Belal Helal, and Deepashree Kedia. code. paper.
A Paxos-Based Strongly Consistent Live Document Editor. Weixin Yu and Kaylee Xie. code. paper.
Reducing Unavailability of Geo-replicated Raft with Automatic Election Timeout Configuration. David Fuziwei Smith and Nick John Riedman. code. paper.
A Minimal Bare-Metal DHCP Server for Ultra-Low-Memory Devices. Linfan Qian, Di Bao, Jihan Xia, and Ruiting Chen. code. paper.
Is SIEVE Better than LRU?. William Hu, Thomas Li, and Balaji M Balachandran. code. paper.
12pm Lunch
1pm
Extending GPU Cluster Schedulers with Fragmentation Awareness. Varun Ramesh, Jiayu Chang, Hyeonggyu Kim, and Stefan Gabriel Ene. code. paper.
ZeRO from Scratch: Evaluating Memory Savings, Communication Overhead, and Bandwidth Sensitivity. Thomason Zhao and Daniel Adkins. code. paper.
DAG-Aware Request Scheduling for LLM Agent Systems Under External Rate Limits. Berk Timothy Yalcinkaya, Filip William Henriksson, and Anushka Roshni De. code. paper.
Bandit-Guided NCCL Tuning for Multi-Node GPU Clusters. Varun Madan and Ashley Yumiko Raigosa. code. paper.
At What Cost? Enhanced Cost Functions for Fair LLM Serving. Karthik Vetrivel, Cary Xiao, Michael Rybalkin, and Jamie Ullman. code. paper.