21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24)
2024-04-16..2024-04-18
Santa Clara, United States
as the organisers put it
The 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) will take place April 16–18, 2024, at the Hyatt Regency Santa Clara in Santa Clara, CA, USA. NSDI focuses on the design principles, implementation, and practical evaluation of networked and distributed systems.
| title | author | length |
|---|---|---|
| Horus: Granular In-Network Task Scheduler for Cloud Datacenters | Parham Yassini, Khaled Diab, Saeed Zangeneh, Mohamed Hefeeda | 18m |
A distributed in-network scheduler that runs entirely in the data plane and is implemented in P4 on Intel Tofino switches, Horus tracks per-worker load with… | ||
| Fast Vector Query Processing for Large Datasets Beyond GPU Memory with Reordered... | Zili Zhang, Fangyue Liu, Gang Huang, Xuanzhe Liu, Xin Jin | 15m |
RUMMY is a GPU-backed vector query engine that pipelines host-to-GPU transfers and computation to handle billion-scale datasets beyond GPU RAM. It implements… | ||
| LoLKV: The Logless, Linearizable, RDMA-based Key-Value Storage System | Ahmed Alquraan, Sreeharsha Udayashankar, Virendra Marathe, Bernard Wong, Samer Al-Kiswany | 14m |
LoLKV implements a logless, linearizable RDMA key-value store that uses one critical-path RDMA write to replicate a packed segment entry and an asynchronous… | ||
| Making Kernel Bypass Practical for the Cloud with Junction | Joshua Fried, Gohar Irfan Chaudhry, Enrique Saurez, Esha Choukse, Íñigo Goiri, Sameh Elnikety, Rodrigo Fonseca, Adam Belay | 16m |
Junction is a kernel-bypass system that actually packs thousands of unmodified Linux instances onto a machine by cutting pinned memory and avoiding constant… | ||
| Sifter: An Inversion-Free and Large-Capacity Programmable Packet Scheduler | Peixuan Gao, Anthony Dalleggio, Jiajin Liu, Chen Peng, Yang Xu, H. Jonathan Chao | 13m |
Sifter claims inversion-free, large-capacity programmable scheduling by combining Sift Sorting coarse buckets (FIFO rotating RCQ) with a small mini PIFO for… | ||
| Flow Scheduling with Imprecise Knowledge | Wenxin Li, Xin He, Yuan Liu, Keqiu Li, Kai Chen, Zhao Ge, Zewei Guan, Heng Qi, Song Zhang, Guyue Liu | 16m |
QCLIMB uses random forest regression forests to produce per-flow lower and upper bounds (over 99.9% of flows are tightly bounded) and runs a low-bound-based… | ||
| Pudica: Toward Near-Zero Queuing Delay in Congestion Control for Cloud Gaming | Shibo Wang, Shusen Yang, Xiao Kong, Chenglei Wu, Longwei Jiang, Chenren Xu, Cong Zhao, Xuesong Yang, Jianjun Xiao, Xin Liu, Changxi Zheng, Jing Wang, Honghao Liu | 19m |
Pudica is an end-to-end congestion control that probes link BUR (bandwidth utilization ratio) by sending paced frame traffic plus occasional payload-free probe… | ||
| Revisiting Congestion Control for Lossless Ethernet | Yiran Zhang, Qingkai Meng, Chaolei Hu, Fengyuan Ren | 18m |
Revisits lossless Ethernet congestion control and proposes ACK-driven congestion control (ACC) that uses receiver ACK time series and packet conservation to… | ||
| Autothrottle: A Practical Bi-Level Approach to Resource Management for SLO-Targeted... | Zibo Wang, Pinghe Li, Chieh-Jan Mike Liang, Feng Wu, Francis Y. Yan | 17m |
A bi-level framework called Autothrottle pairs lightweight per-service controllers (Captain) with an application-level learner (Tower) that outputs throttle… | ||
| Jolteon: Unleashing the Promise of Serverless for Serverless Workflows | Zili Zhang, Chao Jin, Xin Jin | 18m |
An orchestrator called Jolteon implements a stochastic performance model combining analytical whitebox formulas and blackbox variability, formulates a… | ||
| Can't Be Late: Optimizing Spot Instance Savings under Deadlines | Zhanghao Wu, Wei-Lin Chiang, Ziming Mao, Zongheng Yang, Eric Friedman, Scott Shenker, Ion Stoica | 18m |
Proposes a parameter-free policy Uniform Progress that switches to on-demand only when observed progress falls below the time-uniform expectation, and… | ||
| Towards Intelligent Automobile Cockpit via A New Container Architecture | Lin Jiang, Feiyu Zhang, Jiang Ming | 17m |
AutoVP is a virtualization architecture that runs non-safety multimedia on a Android container on a consumer SoC while offloading real-time safety tasks to an… | ||
| MuCache: A General Framework for Caching in Microservice Graphs | Haoran Zhang, Konstantinos Kallas, Spyros Pavlatos, Rajeev Alur, Sebastian Angel, Vincent Liu | 14m |
MuCache is a framework that injects per-service cache managers and sidecar wrappers to serve hits locally without blocking requests, relying on a non-blocking… | ||
| A large-scale deployment of DCTCP | Abhishek Dhamija, Balasubramanian Madhavan, Hechao Li, Jie Meng, Shrikrishna Khare, Madhavi Rao, Lawrence Brakmo, Neil Spring, Prashanth Kannan, Srikanth Sundaresan, Soudeh Ghorbani | 14m |
DCTCP rollout across thousands of racks reduces retransmissions by about 75% drop in retransmissions and cuts one latency-sensitive service’s P90/P99 by 38%… | ||
| TECC: Towards Efficient QUIC Tunneling via Collaborative Transmission Control | Jiaxing Zhang, Furong Yang, Ting Liu, Qinghua Wu, Wu Zhao, Yuanbo Zhang, Wentao Chen, Yanmei Liu, Hongyu Guo, Yunfei Ma, Zhenyu Li | 14m |
TECC implements a feedback framework where the tunnel server sends RTT, queue length, retransmit rate and pacing to the client so the end-to-end QUIC sender… | ||
| iStack: A General and Stateful Name-based Protocol Stack for Named Data Networking | Tianlong Li, Tian Song, Yating Yang | 18m |
iStack implements a Linux kernel NDN stack (4.14/4.19) with a name resolution + binding-prefix table, a two-layer face system, socket-compatible APIs and… | ||
| Cloudcast: High-Throughput, Cost-Aware Overlay Multicast in the Cloud | Sarah Wooders, Shu Liu, Paras Jain, Xiangxi Mo, Joseph E. Gonzalez, Vincent Liu, Ion Stoica | 18m |
Cloudcast builds an overlay multicast system that spawns ephemeral waypoints (VM routers) and emits per-VM router programs to execute striped, multi-tree… | ||
| Understanding Routable PCIe Performance for Composable Infrastructures | Wentao Hou, Jie Zhang, Zeke Wang, Ming Liu | 17m |
rPCIeBench is a software–hardware co-designed benchmark that implements four DMA primitives to measure routable PCIe paths using a Gig Fabric X RS4024 switch… | ||
| Alea-BFT: Practical Asynchronous Byzantine Fault Tolerance | Diogo S. Antunes, Afonso N. Oliveira, André Breda, Matheus Guilherme Franco, Henrique Moniz, Rodrigo Rodrigues | 15m |
Alea-BFT is an asynchronous BFT state-machine protocol that reintroduces a per-request designated replica and a two-stage pipeline: an efficient leader-led… | ||
| Harmony: A Congestion-free Datacenter Architecture | Saksham Agarwal, Qizhe Cai, Rachit Agarwal, David Shmoys, Amin Vahdat | 20m |
Claims a datacenter architecture that guarantees bounded queueing at each switch so messages once sent observe a worst‑case delay and no congestion‑related… | ||
| SwiftPaxos: Fast Geo-Replicated State Machines | Fedor Ryabinin, Alexey Gotsman, Pierre Sutra | 16m |
Protocol executes a command in 2 message delays when there is no contention and in 3 delays otherwise, by letting replicas vote twice — once for a local… | ||
| The Bedrock of Byzantine Fault Tolerance: A Unified Platform for BFT Protocols... | Mohammad Javad Amiri, Chenyuan Wu, Divyakant Agrawal, Amr El Abbadi, Boon Thau Loo, Mohammad Sadoghi | 17m |
Bedrock is a unified platform that encodes BFT protocols as points in a 20 design dimensions space, implements them from a domain-specific language, and runs… | ||
| DINT: Fast In-Kernel Distributed Transactions with eBPF | Yang Zhou, Xingyu Xiang, Matthew Kiley, Sowmya Dharanipragada, Minlan Yu | 19m |
DINT implements eBPF offload of frequent-path transaction operations into the kernel (2K lines of eBPF plus 4K C++) using kernel maps, per-core ring buffers… | ||
| Brain-on-Switch: Towards Advanced Intelligent Network Data Plane via NN-Driven Traffic... | Jinzhu Yan, Haotian Xu, Zhuotao Liu, Qi Li, Ke Xu, Mingwei Xu, Jianping Wu | 14m |
Implements a prototype on P4 Barefoot Tofino and claims line‑speed neural inference by running a binarized RNN with sliding‑window recurrence entirely on the… | ||
| The Eternal Tussle: Exploring the Role of Centralization in IPFS | Yiluo Wei, Dennis Trautwein, Yiannis Psaras, Ignacio Castro, Will Scott, Aravindh Raman, Gareth Tyson | 14m |
IPFS adds three centralized components and measures concrete gains and costs: IPNI indexer accepts massive publication and holds ~100× more provider records… | ||
| BBQ: A Fast and Scalable Integer Priority Queue for Hardware Packet Scheduling | Nirav Atre, Hugo Sadok, Justine Sherry | 18m |
BBQ implements a pipeline-parallel integer priority queue (hierarchical find-first-set) in hardware, synthesized on a commodity FPGA and projected to a 7nm… | ||
| Sirius: Composing Network Function Chains into P4-Capable Edge Gateways | Jiaqi Gao, Jiamin Cao, Yifan Li, Mengqi Liu, Ming Tang, Dennis Cai, Ennan Zhai | 16m |
System composes dozens of distinct NF chains into a single P4 program, synthesizing traffic identification tables, pipelining loops via recirculations, and… | ||
| Empower Programmable Pipeline for Advanced Stateful Packet Processing | Yong Feng, Zhikang Chen, Haoyu Song, Yinchao Zhang, Hanyi Zhou, Ruoyu Sun, Wenkuo Dong, Peng Lu, Shuxin Liu, Chuwen Zhang, Yang Xu, Bin Liu | 14m |
Proposes RAPID: a pipeline extension that adds a low-cost side ring for fast backward state writes and uses speculative execution to avoid stalling, plus an… | ||
| GRACE: Loss-Resilient Real-Time Video through Neural Codecs | Yihua Cheng, Ziyi Zhang, Hanchen Li, Anton Arapin, Yue Zhang, Qizheng Zhang, Yuhan Liu, Kuntai Du, Xu Zhang, Francis Y. Yan, Amrita Mazumdar, Nick Feamster, Junchen Jiang | 16m |
A neural video codec that jointly trains encoder and decoder under simulated packet losses and randomly zeroed tensor partitions yields graceful quality… | ||
| LiFteR: Unleash Learned Codecs in Video Streaming with Loose Frame Referencing | Bo Chen, Zhisheng Yan, Yinjie Zhang, Zhe Yang, Klara Nahrstedt | 17m |
LiFteR replaces tight, sequential frame references with LFR (loose frame referencing) and a binary-tree referencing schedule, letting a learned DNN codec… | ||
| MadEye: Boosting Live Video Analytics Accuracy with Adaptive Camera Configurations | Mike Wong, Murali Ramanujam, Guha Balakrishnan, Ravi Netravali | 17m |
MadEye is a camera–server system that actively reorients commodity PTZ cameras, runs a fast search over rotations/zooms, and uses on‑camera lightweight… | ||
| Gemino: Practical and Robust Neural Compression for Video Conferencing | Vibhaalakshmi Sivaraman, Pantea Karimi, Vedantha Venkatapathy, Mehrdad Khani, Sadjad Fouladi, Mohammad Alizadeh, Frédo Durand, Vivienne Sze | 17m |
System reconstructs 1024x1024 video by sending tiny downsampled frames plus a single high-resolution reference and performing high-frequency-conditional… | ||
| ARTEMIS: Adaptive Bitrate Ladder Optimization for Live Video Streaming | Farzad Tashtarian, Abdelhak Bentaleb, Hadi Amirpour, Sergey Gorinsky, Junchen Jiang, Hermann Hellwagner, Christian Timmerer | 14m |
System implements a cloud service that advertises a mega manifest of many candidate bitrate-resolution pairs, collects CDN logs and encoder quality metrics… | ||
| Credence: Augmenting Datacenter Switch Buffer Sharing with ML Predictions | Vamsi Addanki, Maciej Pacut, Stefan Schmid | 16m |
CREDENCE augments commodity drop-tail buffers with ML predictions to emulate push-out behavior, using per-queue thresholds and an “arithmetic push-out” trick… | ||
| Seer: Enabling Future-Aware Online Caching in Networked Systems | Jason Lei, Vishal Shrivastav | 18m |
Seer claims to shrink cache misses up to 65% and flow completion time up to 78% versus LRU by sending small state-index notifications ahead of packets and… | ||
| Reverie: Low Pass Filter-Based Switch Buffer Sharing for Datacenters with RDMA and TCP... | Vamsi Addanki, Wei Bai, Stefan Schmid, Maria Apostolaki | 15m |
Reverie replaces the vendor Sonic model with a single shared pool, accounting each packet once and consolidating ingress and egress views to avoid… | ||
| Precise Data Center Traffic Engineering with Constrained Hardware Resources | Shawn Shuoshuo Chen, Keqiang He, Rui Wang, Srinivasan Seshan, Peter Steenkiste | 17m |
System adapts switch memory allocation to flow volume and precision loss, prunes select ports and merges identical weight distributions to implement TE on… | ||
| Multitenant In-Network Acceleration with SwitchVM | Sajy Khashab, Alon Rashelbach, Mark Silberstein | 17m |
Proposes SwitchVM: a sandboxed VM that runs tenant Turing-complete DPFs on programmable switches, implemented as an interpreter built in P4 on Intel Tofino… | ||
| Characterization of Large Language Model Development in the Datacenter | Qinghao Hu, Zhisheng Ye, Zerui Wang, Guoteng Wang, Meng Zhang, Qiaoling Chen, Peng Sun, Dahua Lin, Xiaolin Wang, Yingwei Luo, Yonggang Wen, Tianwei Zhang | 17m |
Six-month datacenter trace from Acme covers >4,000 A100 GPUs and ~1M jobs (7B–123B model family, 60% GPU jobs) and shows most GPU time coming from large jobs… | ||
| QuickUpdate: a Real-Time Personalization System for Large-Scale Recommendation Models | Kiran Kumar Matam, Hani Ramezani, Fan Wang, Zeliang Chen, Yue Dong, Maomao Ding, Zhiwei Zhao, Zhengyu Zhang, Ellie Wen, Assaf Eisenman | 12m |
Real-time personalization for terabyte-scale recommendation embeddings reduces published bandwidth by over 13x while matching a fully fresh model’s accuracy… | ||
| MegaScale: Scaling Large Language Model Training to More Than 10,000 GPUs | Ziheng Jiang, Haibin Lin, Yinmin Zhong, Qi Huang, Yangrui Chen, Zhi Zhang, Yanghua Peng, Xiang Li, Cong Xie, Shibiao Nong, Yulu Jia, Sun He, Hongmin Chen, Zhihao Bai, Qi Hou, Shipeng Yan, Ding Zhou, Yiyao Sheng, Zhuo Jiang, Haohan Xu, Haoran Wei, Zhang Zhang, Pengfei Nie, Leqi Zou, Sida Zhao, Liang Xiang, Zherui Liu, Zhe Li, Xiaoying Jia, Jianxi Ye, Xin Jin, Xin Liu | 16m |
MegaScale is a production system that trains LLMs on more than 10,000 GPUs and achieves concrete gains: 55.2% Model FLOPs Utilization when training a 175B… | ||
| Resiliency at Scale: Managing Google’s TPUv4 Machine Learning Supercomputer | Yazhou Zu, Alireza Ghaffarkhah, Hoang-Vu Dang, Brian Towles, Steven Hand, Safeen Huda, Adekunle Bello, Alexander Kolbasov, Arash Rezaei, Dayou Du, Steve Lacy, Hang Wang, Aaron Wisner, Chris Lewis, Henri Bahini | 18m |
Describes a production 4096-node accelerator cluster built as 4096-node TPUv4 supercomputers with custom 3D torus-like fabric and per-chip ICI switches… | ||
| NN-Defined Modulator: Reconfigurable and Portable Software Modulator on IoT Gateways | Jiazhao Wang, Wenchao Jiang, Ruofeng Liu, Bin Hu, Demin Gao, Shuai Wang | 16m |
Neural-network models replace handcoded modulators so an IoT gateway compiles a modulator once and runs it across heterogeneous boards via ONNX, with a… | ||
| Democratizing Direct-to-Cell Low Earth Orbit Satellite Networks | Lixin Liu, Yuanjie Li, Hewu Li, Jiabo Yang, Wei Liu, Jingyi Lan, Yufeng Wang, Jiarui Li, Jianping Wu, Qian Wu, Jun Liu, Zeqi Lai | 18m |
Paper claims MOSAIC replaces fragile hop-by-hop cellular sessions with SIM-based one-time tokens that let satellites locally authorize pay-as-you-go 4G/5G… | ||
| Known Knowns and Unknowns: Near-realtime Earth Observation Via Query Bifurcation in... | Bill Tao, Om Chabra, Ishani Janveja, Indranil Gupta, Deepak Vasisht | 17m |
Serval is a system that splits queries into glacial/dynamic filter bifurcation, precomputes stable filters on ground stations and runs only rapidly changing… | ||
| Spectrumize: Spectrum-efficient Satellite Networks for the Internet of Things | Vaibhav Singh, Tusher Chakraborty, Suraj Jog, Om Chabra, Deepak Vasisht, Ranveer Chandra | 18m |
Spectrumize is a system for downlink from LEO pico-satellites that uses predicted Doppler signature sequences as CDMA-like codes to correlate an entire… | ||
| Application-Level Service Assurance with 5G RAN Slicing | Arjun Balasingam, Manikanta Kotaru, Paramvir Bahl | 17m |
Zipper provides app-level 5G RAN guarantees by casting slice bandwidth allocation as a model predictive controller that iteratively invokes slice MACs and uses… | ||
| CHISEL: An optical slice of the wide-area network | Abhishek Vijaya Kumar, Bill Owens, Nikolaj Bjørner, Binbin Guan, Yawei Yin, Paramvir Bahl, Rachee Singh | 18m |
CHISEL dynamically carves optical slices across a cloud WAN, implemented as a MILP-based allocator that maps slice requests to contiguous spectrum and… | ||
| LuoShen: A Hyper-Converged Programmable Gateway for Multi-Tenant Multi-Service Edge... | Tian Pan, Kun Liu, Xionglie Wei, Yisong Qiao, Jun Hu, Zhiguo Li, Jun Liang, Tiesheng Cheng, Wenqiang Su, Jie Lu, Yuke Hong, Zhengzhong Wang, Zhi Xu, Chongjing Dai, Peiqiao Wang, Xuetao Jia, Jianyuan Lu, Enge Song, Jun Zeng, Biao Lyu, Ennan Zhai, Jiao Zhang, Tao Huang, Dennis Cai, Shunmin Zhu | 14m |
LuoShen consolidates a full VPC network into a 2U server-switch with a P4-centric pipeline, using pipeline folding and meticulous on-chip resource budgeting to… | ||
| Sprinter: Speeding Up High-Fidelity Crawling of the Modern Web | Ayush Goel, Jingyuan Zhu, Ravi Netravali, Harsha V. Madhyastha | 17m |
Sprinter combines a small browser-crawled subset with a browserless phase that replays results from a computation cache built by rewriting JS and recording… | ||
| Hairpin: Rethinking Packet Loss Recovery in Edge-based Interactive Video Streaming | Zili Meng, Xiao Kong, Jing Chen, Bo Wang, Mingwei Xu, Rui Han, Honghao Liu, Venkat Arun, Hongxin Hu, Xue Wei | 17m |
Hairpin rethinks loss recovery for edge-based interactive video by choosing per-round mixes of original packets, retransmissions, and redundancy rather than a… | ||
| Finding Adversarial Inputs for Heuristics using Multi-level Optimization | Pooria Namyar, Behnaz Arzani, Ryan Beckett, Santiago Segarra, Himanshu Raj, Umesh Krishnaswamy, Ramesh Govindan, Srikanth Kandula | 17m |
MetaOpt is a solver-backed tool that encodes heuristics and an optimal baseline as a bilevel optimization, automatically rewrites inner problems into… | ||
| Towards provably performant congestion control | Anup Agarwal, Venkat Arun, Devdeep Ray, Ruben Martins, Srinivasan Seshan | 17m |
Framework creates a belief set that exhaustively maps observed send/ACK time series to possible bottleneck bandwidth, buffer, and delay states, and uses… | ||
| EPVerifier: Accelerating Update Storms Verification with Edge-Predicate | Chenyang Zhao, Yuebin Guo, Jingyu Wang, Qi Qi, Zirui Zhuang, Haifeng Sun, Lingqi Guo, Yuming Xie, Jianxin Liao | 14m |
EPVerifier replaces global equivalence-class models with an edge-predicate model (one BDD-like predicate per link), so only actually affected edges are updated… | ||
| Netcastle: Network Infrastructure Testing At Scale | Rob Sherwood, Jinghao Shi, Ying Zhang, Neil Spring, Srikanth Sundaresan, Jasmeet Bagga, Prathyusha Peddi, Vineela Kukkadapu, Rashmi Shrivastava, Manikantan KR, Pavan Patil, Srikrishna Gopu, Varun Varadan, Ethan Shi, Hany Morsy, Yuting Bu, Renjie Yang, Rasmus Jönsson, Wei Zhang, Jesus Jussepen Arredondo, Diana Saha, Sean Choi | 19m |
Netcastle provides CI/CD network testing as a service that automates topology setup, resource checkout, retries and health checks and runs 500K+ tests per day… | ||
| MESSI: Behavioral Testing of BGP Implementations | Rathin Singha, Rajdeep Mondal, Ryan Beckett, Siva Kesava Reddy Kakarla, Todd Millstein, George Varghese | 18m |
MESSI is a model-based tester that symbolically generates joint configuration–state–route triples to exercise BGP behavior, producing about 150K tests and… | ||
| A High-Performance Design, Implementation, Deployment, and Evaluation of The Slim Fly... | Nils Blach, Maciej Besta, Daniele De Sensi, Jens Domke, Hussein Harake, Shigang Li, Patrick Iff, Marek Konieczny, Kartik Lakhotia, Ales Kubicek, Marcel Ferrari, Fabrizio Petrini, Torsten Hoefler | 19m |
Builds and validates the first real-world Slim Fly cluster: a 200-server, 50-router installation with practical cabling patterns and an InfiniBand routing… | ||
| Crescent: Emulating Heterogeneous Production Network at Scale | Zhaoyu Gao, Anubhavnidhi Abhashkumar, Zhen Sun, Weirong Jiang, Yi Wang | 17m |
Crescent emulates production networks by running switch vendor images inside containers and links them with veth pairs and OVS VXLAN, trading full-network… | ||
| Reasoning about Network Traffic Load Property at Production Scale | Ruihan Li, Fangdan Ye, Yifei Yuan, Ruizhen Yang, Bingchuan Tian, Tianchen Guo, Hao Wu, Xiaobo Zhu, Zhongyu Guan, Qing Ma, Xianlong Zeng, Chenren Xu, Dennis Cai, Ennan Zhai | 17m |
NSDI '24 - Reasoning about Network Traffic Load Property at Production Scale
Ruihan Li, Peking University and Alibaba Cloud; Fangdan Ye, Yifei Yuan, Ruizhen… | ||
| POSEIDON: A Consolidated Virtual Network Controller that Manages Millions of Tenants... | Biao Lyu, Enge Song, Tian Pan, Jianyuan Lu, Shize Zhang, Xiaoqing Sun, Lei Gao, Chenxiao Wang, Han Xiao, Yong Pan, Xiuheng Chen, Yandong Duan, Weisheng Wang, Jinpeng Long, Yanfeng Wang, Kunpeng Zhou, Zhigang Zong, Xing Li, Guangwang Li, Pengyu Zhang, Peng Cheng, Jiming Chen, Shunmin Zhu | 17m |
Poseidon is a consolidated virtual network controller deployed on Alibaba Cloud for 3+ years that replaces 50+ service controllers with a unified… | ||
| OPPerTune: Post-Deployment Configuration Tuning of Services Made Easy | Gagan Somashekar, Karan Tandon, Anush Kini, Chieh-Chun Chang, Petr Husak, Ranjita Bhagwan, Mayukh Das, Anshul Gandhi, Nagarajan Natarajan | 18m |
OpperTune is a deployed service that automatically picks which of hundreds of config knobs to tune, chooses whether to share or split tuning instances across… | ||
| Parcae: Proactive, Liveput-Optimized DNN Training on Preemptible Instances | Jiangfei Duan, Ziang Song, Xupeng Miao, Xiaoli Xi, Dahua Lin, Harry Xu, Minjia Zhang, Zhihao Jia | 16m |
Parcae is a system that proactively adapts hybrid data+pipeline parallelization to predicted spot-instance changes, using a liveput metric that balances… | ||
| Accelerating Neural Recommendation Training with Embedding Scheduling | Chaoliang Zeng, Xudong Liao, Xiaodian Cheng, Han Tian, Xinchen Wan, Hao Wang, Kai Chen | 15m |
A runtime embedding scheduler called Herald decides where embeddings are trained and which to sync, using an adaptive location-aware inputs allocator and an… | ||
| DISTMM: Accelerating Distributed Multimodal Model Training | Jun Huang, Zhen Zhang, Shuai Zheng, Feng Qin, Yida Wang | 18m |
DISTIMM implements per-submodule parallelism choices (tensor, data, hybrid), adaptive partitioning and non‑collocated placement to balance M/K/N dimensions and… | ||
| Approximate Caching for Efficiently Serving Text-to-Image Diffusion Models | Shubham Agarwal, Subrata Mitra, Sarthak Chakraborty, Srikrishna Karanam, Koyel Mukherjee, Shiv Kumar Saini | 17m |
Caches and reuses intermediate diffusion noise states to skip early sampling steps, claiming 21% GPU compute savings, 19.8% end-to-end latency reduction, and… | ||
| THC: Accelerating Distributed Deep Learning Using Tensor Homomorphic Compression | Minghao Li, Ran Ben Basat, Shay Vargaftik, ChonLam Lao, Kevin Xu, Michael Mitzenmacher, Minlan Yu | 16m |
Proposes Tensor Homomorphic Compression that makes compressed gradients directly summable (homomorphic property) so parameter servers or programmable switches… | ||
| Accelerating Skewed Workloads With Performance Multipliers in the TurboDB Distributed... | Jennifer Lam, Jeffrey Helt, Wyatt Lloyd, Haonan Lu | 16m |
Build replaces one CockroachDB server with a Cicada turbo node and promotes the 40 million hottest keys to it, implementing Hybrid Concurrency Control to… | ||
| SIEVE is Simpler than LRU: an Efficient Turn-Key Eviction Algorithm for Web Caches | Yazhuo Zhang, Juncheng Yang, Yao Yue, Ymir Vigfusson, K.V. Rashmi | 18m |
SIEVE is an eviction algorithm that uses a moving “S” hand to give lazy promotion and quick demotion, keeping hits lock-free so cache hits require no locking… | ||
| Harvesting Idle Memory for Application-managed Soft State with Midas | Yifan Qiao, Zhenyuan Ruan, Haoran Ma, Adam Belay, Miryung Kim, Harry Xu | 18m |
Midas implements a new soft memory abstraction plus a userlinked runtime and minimal kernel coordinator to let applications place evictable objects in idle RAM… | ||
| Efficient Exposure of Partial Failure Bugs in Distributed Systems with Inferred... | Haoze Wu, Jia Pan, Peng Huang | 18m |
Legolas statically instruments code with fine-grained injection hooks and automatically infers abstract state variables (ASVs) to compress millions of fault… | ||
| Load is not what you should balance: Introducing Prequal | Bartek Wydrowski, Robert Kleinberg, Stephen M. Rumble, Aaron Archer | 19m |
PReQuaL is a load balancer that actively probes servers and uses the power-of-d-choices with asynchronous reusable probes to pick based on estimated latency… | ||
| Orthcatter: High-throughput In-band OFDM Backscatter with Over-the-Air Code Division | Caihui Du, Jihong Yu, Rongrong Zhang, Ju Ren, Jianping An | 19m |
Orthcatter claims a working prototype that turns a Wi‑Fi OFDM exciter into a carrier for high‑rate in‑band backscatter, reaching 248 kbps and BER ≈ 10^-4 while… | ||
| EdgeRIC: Empowering Real-time Intelligent Optimization and Control in NextG Cellular... | Woo-Hyun Ko, Ushasi Ghosh, Ujwal Dinesha, Raini Wu, Srinivas Shakkottai, Dinesh Bharadia | 17m |
A real-time RIC called EdgeRIC is implemented co‑located with the DU (built on SRS RAN) and synchronized to the TTI with 1 ms TTI sync, using a lightweight… | ||
| ADR-X: ANN-Assisted Wireless Link Rate Adaptation for Compute-Constrained Embedded... | Hao Yin, Murali Ramanujam, Joe Schaefer, Stan Adermann, Srihari Narlanka, Perry Lea, Ravi Netravali, Krishna Chintalapudi | 17m |
System achieves 10× lower packet loss (down to 0.1–0.5% versus typical 2–10%) while running ~100× faster than prior RL work, with inference taking 170–380 μs… | ||
| RFID+: Spatially Controllable Identification of UHF RFIDs via Controlled Magnetic Fields | Donghui Dai, Zhenlin An, Zheng Gong, Qingrui Pan, Lei Yang | 16m |
RFID+ builds a magnetically driven UHF RFID system that powers off-the-shelf tags via their matching loops using a capacitor-segmented multi-turn coil and a… | ||
| SMUFF: Towards Line Rate Wi-Fi Direct Transport with Orchestrated On-device Buffer... | Chengke Wang, Hao Wang, Yuhan Zhou, Yunzhe Ni, Feng Qian, Chenren Xu | 17m |
SMUFF implements orchestrated on-device buffer management that monitors every buffer on the sender–receiver path and dynamically backlogs just enough data in… | ||
| Vulcan: Automatic Query Planning for Live ML Analytics | Yiwen Zhang, Xumiao Zhang, Ganesh Ananthanarayanan, Anand Iyer, Yuanchao Shu, Victor Bahl, Z. Morley Mao, Mosharaf Chowdhury | 16m |
Vulcan automatically generates full query plans for live ML analytics by building pipelines, deciding filter order, and jointly optimizing operator placement… | ||
| CASSINI: Network-Aware Job Scheduling in Machine Learning Clusters | Sudarsanan Rajasekaran, Manya Ghobadi, Aditya Akella | 16m |
CASSINI introduces a network-aware job scheduler that packs ML jobs by their periodic traffic using a geometric abstraction (circles and a unified LCM-based… | ||
| Towards Domain-Specific Network Transport for Distributed DNN Training | Hao Wang, Han Tian, Jingrong Chen, Xinchen Wan, Jiacheng Xia, Gaoxiong Zeng, Wei Bai, Junchen Jiang, Yong Wang, Kai Chen | 19m |
MLT is a domain-specific transport for distributed DNN training that implements a 3-step progressive scheme: (1) per-packet load balancing across multiple… | ||
| Swing: Short-cutting Rings for Higher Bandwidth Allreduce | Daniele De Sensi, Tommaso Bonato, David Saam, Torsten Hoefler | 19m |
New algorithm Swing shortens allreduce paths by short-cutting rings and swinging between torus directions, splitting each message so every node uses all local… | ||
| LitePred: Transferable and Scalable Latency Prediction for Hardware-Aware Neural... | Chengquan Feng, Li Lyna Zhang, Yuanchi Liu, Jiahang Xu, Chengruidong Zhang, Zhiyuan Wang, Ting Cao, Mao Yang, Haisheng Tan | 11m |
LitePred claims it predicts DNN inference latency on new edge platforms by transferring existing predictors rather than rebuilding them. It uses a VAE data… | ||
| Harmonic: Hardware-assisted RDMA Performance Isolation for Public Clouds | Shin haong | ? |
Claims a deployable hardware–software design that enforces RDMA performance isolation by measuring and throttling at the PCIe boundary. Implements a… | ||
| LDB: An Efficient Latency Profiling Tool for Multithreaded Applications | Inho Cho, Seo Jin Park, Ahmed Saeed, Mohammad Alizadeh, Adam Belay | 20m |
LDB is a software-only profiler that samples call stacks using a background thread and records request and synchronization tags to attribute tail latency to… | ||
| UFO: The Ultimate QoS-Aware Core Management for Virtualized and Oversubscribed Public... | Yajuan Peng, Shuang Chen, Yi Zhao, Zhibin Yu | 15m |
UFO is a feedback controller that coordinates host and guest cores, monitors per-VM scheduling frequency every second, predicts when extra pCPUs are needed and… | ||
| Automatic Parallelization of Software Network Functions | Francisco Pereira, Fernando M. V. Ramos, Luis Pedrosa | 16m |
Maestro analyzes sequential NF code with exhaustive symbolic execution and static analysis, encodes sharding constraints as SMT queries, and outputs NIC RSS… | ||
| AutoSketch: Automatic Sketch-Oriented Compiler for Query-driven Network Telemetry | Haifeng Sun, Qun Huang, Jinbo Sun, Wei Wang, Jiaheng Li, Fuliang Li, Yungang Bao, Xin Yao, Gong Zhang | 15m |
AutoSketch automatically compiles high-level query operators into sketch instances, then searches configurations to meet a user-specified accuracy intent… | ||
| Leo: Online ML-based Traffic Classification at Multi-Terabit Line Rate | Syed Usman Jafri, Sanjay Rao, Vishal Shrivastav, Mohit Tawarmalani | 18m |
System implements decision trees in the switch data plane and claims packet-by-packet classification at multi‑terabit line rate on Intel Tofino by combining a… | ||
| Sequence Abstractions for Flexible, Line-Rate Network Monitoring | Andrew Johnson, Ryan Beckett, Xiaoqi Chen, Ratul Mahajan, David Walker | 16m |
A new system called FLM compiles sequence-monitoring programs to P4 that run on programmable switch ASICs, splitting specs into state management plus a regular… | ||
| OctoSketch: Enabling Real-Time, Continuous Network Monitoring over Multiple Cores | Yinda Zhang, Peiqing Chen, Zaoxing Liu | 14m |
OctoSketch scales sketches across many cores by sending only counters that change enough to matter, using continuous change-based aggregation. It strips heavy… | ||
| NR-Surface: NextG-ready µW-reconfigurable mmWave Metasurface | Minseok Kim, Namjo Ahn, Song Min Kim | 16m |
Claims a fully NR-compliant, real-time reconfigurable mmWave metasurface that runs at ~242.7 µW end-to-end and lasts ~2.1 years on one AA battery, using a 15… | ||
| Cyclops: A Nanomaterial-based, Battery-Free Intraocular Pressure (IOP) Monitoring... | Liyao Li, Bozhao Shang, Yun Wu, Jie Xiong, Xiaojiang Chen, Yaxiong Xie | 15m |
Cyclops is a battery-free contact lens that embeds a three-layer antenna (two metal layers with a nanomaterial sensing layer) and an RFID chip with a… | ||
| Habitus: Boosting Mobile Immersive Content Delivery through Full-body Pose Tracking and.. | Anlan Zhang, Chendong Wang, Yuming Hu, Ahmad Hassan, Zejun Zhang, Bo Han, Feng Qian, Shichang Xu | 16m |
Habitus is a user-space framework that combines full-body pose tracking and multipath networking (omnidirectional + mmWave) to predict mmWave throughput and… | ||
| BFMSense: WiFi Sensing Using Beamforming Feedback Matrix | Enze Yi, Dan Wu, Jie Xiong, Fusang Zhang, Kai Niu, Wenwei Li, Daqing Zhang | 12m |
Paper shows that the beamforming feedback matrix (BFM) from new Wi‑Fi cards carries usable channel information and derives a mathematical relation to CSI, then… | ||
| mmComb: High-speed mmWave Commodity WiFi Backscatter | Yoon Chae, Zhenzhe Lin, Kang Min Bae, Song Min Kim, Parth Pathak | 17m |
mmComb turns 802.11ad/ay beamforming control frames into a backscatter carrier by flipping DBPSK phase to encode tag bits and uses receive beamforming plus a… | ||
| Where The Wild Things Are: Brute-Force SSH Attacks In The Wild And How To Stop Them | Sachin Kumar Singh, Shreeman Gautam, Cameron Cartier, Sameer Patil, Robert Ricci | 15m |
Study analyzes sshd logs from over 500 production servers across four years, recording 381 million failed attempts and peaks of 3.5 million/day. It finds 64%… | ||
| A System to Detect Forged-Origin BGP Hijacks | Thomas Holterbach, Thomas Alfroy, Amreesh Phokeer, Alberto Dainotti, Cristel Pelsser | 16m |
A deployed system called DFOH detects forged-origin BGP hijacks across the global RIB by classifying new AS links with a random-forest model trained on 11… | ||
| NetVigil: Robust and Low-Cost Anomaly Detection for East-West Data Center Security | Kevin Hsieh, Mike Wong, Santiago Segarra, Sathiya Kumaran Mani, Trevor Eberl, Anatoliy Panasyuk, Ravi Netravali, Ranveer Chandra, Srikanth Kandula | 17m |
NetVigil uses flow-level logs aggregated into IP-granularity communication graphs (300 vertices vs ~139k flow-vertex graphs) and a GNN autoencoder trained with… | ||
| TANGO: Secure Collaborative Route Control across the Public Internet | Henry Birge-Lee, Sophia Yoo, Benjamin Herber, Jennifer Rexford, Maria Apostolaki | 14m |
TANGO demonstrates that coordinated edge networks can uncover and use hidden BGP paths: bgp Pathfinder finds a median of three paths (up to 10–12) for many… | ||
| Sidekick: In-Network Assistance for Secure End-to-End Transport Protocols | Gina Yuan, Matthew Sotoudeh, David K. Zhang, Michael Welzl, David Mazières, Keith Winstein | 17m |
Sidekick exposes concise side-channel signalling from a middlebox to an endpoint so encrypted transports like QUIC can get in-network hints without changing… | ||
| VILAM: Infrastructure-assisted 3D Visual Localization and Mapping for Autonomous Driving | Jiahe Cui, Shuyao Shi, Yuze He, Jianwei Niu, Guoliang Xing, Zhenchao Ouyang | 15m |
VILAM claims decimeter-level localization and mapping accuracy using only consumer cameras by fusing roadside LiDAR measurements as global references. It… | ||
| Catch Me If You Can: Laser Tethering with Highly Mobile Targets | Charles J. Carver, Hadleigh Schwartz, Qijia Shao, Nicholas Shade, Joseph Lazzaro, Xiaoxin Wang, Jifeng Liu, Eric Fossum, Xia Zhou | 17m |
Lasertag is a prototype system that fuses a steerable laser and image sensor via a beam splitter and fisheye lens, uses a MEMS mirror and retroreflective… | ||
| MobileConfig: Remote Configuration Management for Mobile Apps at Hyperscale | Matt Guo, Soteris Demetriou, Joey Yang, Michael Leighton, Diedi Hu, Tong Bao, Amit Adhikari, Thawan Kooburat, Annie Kim, Chunqiang Tang | 18m |
MobileConfig runs a cross-platform client and a single server evaluation engine that manages configs for billions of devices since 2015, handling apps with… | ||
| Passengers' Safety Matters: Experiences of Deploying a Large-Scale Indoor Delivery... | Xiubin Fan, Zhongming Lin, Yuming Hu, Tianrui Jiang, Feng Qian, Zhimeng Yin, S.-H. Gary Chan, Dapeng Wu | 17m |
Built a real-time monitoring system that attaches IMU and BLE beacons to delivery trolleys, uses an adapted RSSI‑to‑distance model plus particle filtering and… | ||
| AUGUR: Practical Mobile Multipath Transport Service for Low Tail Latency in Real-Time... | Yuhan Zhou, Tingfeng Wang, Liying Wang, Nian Wen, Rui Han, Jing Wang, Chenglei Wu, Jiafeng Chen, Longwei Jiang, Shibo Wang, Honghao Liu, Chenren Xu | 17m |
AUGUR is a server-side multipath transport service that builds per-session state probability models and solves an Integer Linear Programming problem to pick… | ||
| Zombie: Middleboxes that Don’t Snoop | Collin Zhang, Zachary DeStefano, Arasu Arun, Joseph Bonneau, Paul Grubbs, Michael Walfish | 13m |
Claims the first practical system implementing zero-knowledge middleboxes, built as Zombie and evaluated on DNS filtering, that cuts client and middlebox… | ||
| Solving Max-Min Fair Resource Allocations Quickly on Large Graphs | Pooria Namyar, Behnaz Arzani, Srikanth Kandula, Santiago Segarra, Daniel Crankshaw, Umesh Krishnaswamy, Ramesh Govindan, Himanshu Raj | 17m |
Converts the iterative max‑min fair solver into a single fast optimization by using epsilon weighting plus a sorting network, and introduces a multipath… | ||
| Cloud-LoRa: Enabling Cloud Radio Access LoRa Networks Using Reinforcement Learning... | Muhammad Osama Shahid, Daniel Koch, Jayaram Raghuram, Bhuvana Krishnaswamy, Krishna Chintalapudi, Suman Banerjee | 16m |
Cloud-LoRa builds a practical CRAN for LoRa that compresses IQ samples with a reinforcement-learning controller (PPO) over a DWT compressor, streams from a… | ||
| Cloudy with a Chance of Cyberattacks: Dangling Resources Abuse on Cloud Platforms | Jens Frieß, Tobias Gattermayer, Nethanel Gelernter, Haya Schulmann, Michael Waidner | 19m |
Passive DNS plus periodic HTML sampling and simple signatures reveal targeted takeovers of cloud-hosted names rather than an IP lottery: researchers analyze… | ||
| CAPA: An Architecture For Operating Cluster Networks With High Availability | Bingzhe Liu, Colin Scott, Mukarram Tariq, Andrew D. Ferguson, Phillipa Gill, Richard Alimi, Omid Alipourfard, Deepak Arulkannan, Virginia Jean Beauregard, Patrick Conner, P. Brighten Godfrey, Xander Lin, Joon Ong, Mayur Patel, Amr Sabaa, Arjun Singh, Alex Smirnov, Manish Verma, Prerepa V Viswanadham, Amin Vahdat | 16m |
Google describes CAPA as a three‑layer architecture that funnels all management changes through a thin regulation funnel that enforces best practices such as… | ||
| NetAssistant: Dialogue Based Network Diagnosis in Data Center Networks | Haopei Wang, Anubhavnidhi Abhashkumar, Changyu Lin, Tianrong Zhang, Xiaoming Gu, Ning Ma, Chang Wu, Songlin Liu, Wei Zhou, Yongbin Dong, Weirong Jiang, Yi Wang | 18m |
NetAssistant is a task‑oriented chatbot deployed in ByteDance data centers for over three years that runs 100+ executable troubleshooting workflows built from… | ||
| Klonet: an Easy-to-Use and Scalable Platform for Computer Networks Education | Tie Ma, Long Luo, Hongfang Yu, Xi Chen, Jingzhao Xie, Chongxi Ma, Yunhan Xie, Gang Sun, Tianxi Wei, Li Chen, Yanwei Xu, Nicholas Zhang | 15m |
Klonet is a container‑based, distributed network emulation platform that uses Docker containers and veth pairs with a controller-plus-agents orchestrator and a… | ||
| ExChain: Exception Dependency Analysis for Root Cause Diagnosis | Ao Li, Shan Lu, Suman Nath, Rohan Padhye, Vyas Sekar | 17m |
ExChain is an online monitor that instruments services with low overhead and links silent, state-changing exceptions to distant failures using… | ||
| nothing matches | ||