NSDI '26
as the organisers put it
The 23rd USENIX Symposium on Networked Systems Design and Implementation (NSDI '26) will take place May 4–6, 2026, at the Hyatt Regency Lake Washington in Renton, WA, USA. NSDI focuses on the design principles, implementation, and practical evaluation of networked and distributed systems.
| title | author | length |
|---|---|---|
| Keynote: The Physics of Thought and the Architecture of Intelligence | Amin Vahdat | 57m |
Maps historical 5x efficiency breaks to modern models and hardware, claims transformers delivered ~5x compute efficiency over LSTMs, and argues specialization… | ||
| UNUM: A New Framework for Network Control | Jiayi Chen, Nihal Sharma, Debajit Chakraborty, Saurabh AGARWAL, Jeffrey Zhou, Aditya Akella, Sanjay Shakkottai | 16m |
A transformer-based state embedder (UNUM) predicts bucketized future network signals rather than exact values, improving prediction accuracy by up to ~50%… | ||
| BURST: Seeking High-performance, Interoperability and Scalability in Soft-RDMA | Huijun Shen, Zelong Yue, Jian Yang, Zhuo Jiang, Lang An, Yulin Chen, Yong Zhang, Luochangqi Ding, Xiaolong Zhong, Zhihong Wang, Jie Ding, Hongyu Wu, Jianxi Ye, Xijin Yin, Xingyu Zhang, Xingyu Guo, Guo Chen | 18m |
BURST is a user-space RDMA stack that implements the Verbs API, uses a DPDK data plane, Intel DSA for offloaded memcpy, and a kernel-assistant to mmap app… | ||
| PolicyCache: Intra-flow Learning in Congestion Control | Han Tian, Han Wang, Wenbo Li, Xudong Liao, Decang Sun, Wenxue Li, Donghui Chen, Bin Huang, Senbo Fu, Junxue Zhang, Dian Shen, Kai Chen | 14m |
PolicyCache claims intra‑flow learning that trains and runs a nonparametric Hoeffding Adaptive Tree (HAT) inside each flow, plus backup exploration and micro… | ||
| FastServe: Iteration-Level Preemptive Scheduling for Large Language Model Inference | Bingyang Wu, Yinmin Zhong, Zili Zhang, Shengyu Liu, Fangyue Liu, Yuanhang Sun, Gang Huang, Xuanzhe Liu, Xin Jin | 14m |
FastServe implements iteration-level preemption for autoregressive LLMs, letting the system preempt after each token instead of running to completion. It… | ||
| SYMI: Efficient Mixture-of-Experts Training via Model and Optimizer State Decoupling | Athinagoras Skiadopoulos, Mark Zhao, Swapnil Gandhi, Thomas Norrie, Shrijeet Mukherjee, Christos Kozyrakis | 14m |
SYMI claims a practical solution: it decouple optimizer placement from expert parameters so optimizer state is static optimizer sharding across nodes while… | ||
| Towards Performance Robustness for Microservices | Divyanshu Saxena, Gaurav Vipat, Jiaxin Lin, Jingbo Wang, Isil Dillig, Sanjay Shakkottai, Aditya Akella | 15m |
Paper claims that run-time performance robustness certificates (PERCs) give statistical upper bounds on tail latency under bounded perturbations and that… | ||
| Pilot Execution: Simulating Failure Recovery In Situ for Production Distributed Systems | Zhenyu Li, Angting Cai, Chang Lou | 16m |
PILOT implements in‑situ dry‑runs of recovery by spawning phantom threads that shadow‑copy state, propagate a parallel context across RPCs, and cache expensive… | ||
| BLADE: Adaptive Wi-Fi Contention Control for Next-Generation Real-Time Communication | Fengqian Guo, Yuhan Zhou, Longwei Jiang, Congcong Miao, Yuxin Liu, Chenren Xu, Hancheng Lu, Chang Wen Chen, Yaxiong Xie, Honghao Liu | 17m |
A nationwide yearlong measurement with hacked AP drivers and server telemetry identifies Wi‑Fi last‑hop packet delivery droughts as the root cause of long‑tail… | ||
| Observability Is Eating Your Cores: Fine-Grained Analysis of Microservice Metrics with... | Alessandro Cornacchia, Theophilus A. Benson, Muhammad Bilal, Marco Canini | 18m |
μView implements an on-accelerator observability data plane that runs streaming IPU-hosted sketches to analyze microservice metrics locally and ship only… | ||
| MirrorNet: High-fidelity and Scalable Network Emulation for Software-defined WAN | Congcong Miao, Yuejie Wang, Jianming Wang, Xuefeng Ji, Guozhi Shan, Sirui Li, Pan Fang, Yanke Zhang, JIALIN LI, Xianneng Zou, Guyue Liu | 21m |
MirrorNet recreates a production SD‑WAN by mirroring controller, data plane, and traffic and claims second‑by‑second fidelity while cutting storage… | ||
| From Bits to Tokens: Knowledge-Driven Generative Communication of Multimodal Data | Xingyu Chen, Zihao Feng, Wuqiong Zhao, Jianrong Ding, Ke Sun, Xinyu Zhang | 14m |
Proposes Knowledge-Driven Communication that sends tokenized semantics instead of bits, using multimodal foundation models and offline LLM-generated task… | ||
| KeepON: Supporting Deterministic Traffic on Standard NICs | Chuanyu Xue, Tianyu Zhang, Andrew Loveless, Song Han | 18m |
Claim: software-only driver forces commodity NICs to behave like hardware schedulers by keeping the NIC transmitting a stream of continuous fixed-size… | ||
| RASC: Enhancing Observability & Programmability in Smart Spaces | Anna Karanika, Kai-Siang Wang, Han-Ting Liang, Shalni Sundram, Indranil Gupta | 17m |
Proposes RASC (Request-Acknowledge-Start-Complete) as a four-point action API that exposes acknowledgement, start, and completion events so automations can… | ||
| Decoding RSSI Compression in RFID: Dynamic RCS Modeling and Tag-Intrinsic Power Metrics.. | Jia Liu, Yifei Ma, Xingyu Chen, Haipeng Dai, He Huang, Zihao Lin, Wei Zheng, Junzhao Du, Guihai Chen | 15m |
Decoding RSSI Compression in RFID: Dynamic RCS Modeling and Tag-Intrinsic Power Metrics for Reliable Backscatter Networks
Jia Liu, Yifei Ma, Xingyu Chen, and… | ||
| BBC: Enabling BLE to Support Bluetooth Classic | Hsun-Wei Cho, Kang G. Shin | 16m |
BBC moves Bluetooth Classic processing into software and uses BLE radios only to transmit and receive raw FSK bits, enabling single‑mode BLE chips to talk to… | ||
| SPLIDT: Partitioned Decision Trees for Scalable Stateful Inference at Line Rate | Murayyiam Parvez, Annus Zulfiqar, Roman Beltiukov, Shir Landau Feibish, Walter Willinger, Arpit Gupta, Muhammad Shahbaz | 17m |
SPLIDT implements partitioned decision trees on switches/SmartNICs by cutting trees into per-subtree partitions, training each subtree on fixed-size packet… | ||
| Morphe: High-Fidelity Generative Video Streaming with Vision Foundation Model | Tianyi Gong, Zijian Cao, Zixing Zhang, Jiangkai Wu, Xinggong Zhang, Shuguang Cui, Fangxin Wang | 17m |
Morphe uses a vision foundation model with joint-trained visual tokenizers, variable-resolution spatiotemporal optimization, a super-resolution upscaler and… | ||
| DroidSpeak: KV Cache Sharing Across Fine-tuned Model Variants | Yuhan Liu, Yuyang Huang, Jiayi Yao, Shaoting Feng, Zhuohan Gu, Kuntai Du, Hanchen Li, Yihua Cheng, Junchen Jiang, Shan Lu, Madan Musuvathi, Esha Choukse | 17m |
DroidSpeak enables KV cache reuse across fine‑tuned variants of the same architecture by profiling model pairs to find a small set of critical layers (about… | ||
| Checkmate: Zero Performance Overhead Model Checkpointing via Network Gradient Replication | Ankit Bhardwaj, Weiyang Wang, Jeremy Carin, Adam Belay, Manya Ghobadi | 17m |
Claims a system that achieves per-iteration checkpointing with zero GPU overhead by multicasting gradients off the training path and applying them on a… | ||
| HybridMesh: A Hardware-software Hybrid Approach for Accelerating Service Mesh Ingress | Myoungsung You, Jaehyun Nam, Minjae Seo, Taejune Park, Seungwon Shin | 15m |
HybridMesh implements an ingress gateway that splits tasks between a SmartNIC and CPU proxy, using an NVIDIA BlueField-2 DPU running DPDK for layer‑7 parsing… | ||
| Di-PS: System-Algorithm Co-Design for Asynchronous and Heterogeneous Cross-cluster... | Shengwei Li, Qiaoling Chen, Zhiquan Lai, Penglong Jiao, Wenwen Qu, Kun Cai, Jiaxing Li, Peng Sun, Xingcheng Zhang, Xiaoge Deng, Dongsheng Li, Kai Lu, Tianwei Zhang | 16m |
Di-PS presents a centralized leader-follower PS architecture plus algorithmic controls that reduce cross-cluster stalls and stabilize asynchronous updates. The… | ||
| Stimpack: An Adaptive Rendering Optimization System for Scalable Cloud Gaming | Jin Heo, Vic Wang, Ketan Bhardwaj, Ada Gavrilovska | 16m |
Stimpack adaptively lowers or raises server-side rendering quality based on network-driven compression so GPU cycles go where they change perceived frames… | ||
| HydraServe: Minimizing Cold Start Latency for Serverless LLM Serving in Public Clouds | Chiheng Lou, Sheng Qi, Chao Jin, Dapeng Nie, Haoran Yang, Yu Ding, Xuanzhe Liu, Xin Jin | 17m |
HydraServe implements pipeline parallelism, stage overlapping, and pipeline consolidation to cut serverless LLM cold starts that can reach 45 seconds (model… | ||
| FalconFS: Distributed File System for Large-Scale Deep Learning Pipeline | Jingwei Xu, Junbin Kang, Mingkai Dong, Mingyu Liu, Lu Zhang, Shaohong Guo, Ziyan Qiu, Mingzhen You, Ziyi Tian, Anqi Yu, Tianhong Ding, Xinwei Hu, Haibo Chen | 14m |
FalconFS replaces client-side metadata caching with a stateless-client architecture that sends full paths to servers and resolves them using hybrid metadata… | ||
| DistVS: Large-scale Vector Search with Compute-Memory Disaggregation | Peiqi Yin, Xiao Yan, Shiyuan Deng, Hui Li, Yifan Zhu, Xiangyu Zhi, Jingqi Mao, Ran Xu, Wenliang Zhang, James Cheng | 18m |
DistVS maps low precision PQ into the compute tier, high precision PQ plus the graph into a memory tier, and full vectors on SSDs, and executes an iterative… | ||
| Lemonshark: Asynchronous DAG-BFT With Early Finality | Michael Yiqing Hu, Alvin Hong Yao Yan, Yihan Yang, Xiang Liu, JIALIN LI | 15m |
Lemonshark is an asynchronous DAG‑BFT that reinterprets the DAG at the transactional level to deliver early finality for blocks before leader commitment by… | ||
| CacheCatalyst: Enhancing Web Caching for the Latency-Constrained Internet | Mohammad Hosseini, Sina Darabi, Hannaneh B. Pasandi, Patrick Eugster, Mahmood Choopani | 23m |
CacheCatalyst validates a whole page cache on the HTML fetch by adding an X-Cache-Resources header from browser and an X-Validated response flag, implemented… | ||
| ZooRoute: Enhancing Cloud-Scale Network Reliability via Candidate Path Provisioning... | Xiaoqing Sun, Xing Li, Xionglie Wei, Tian Pan, Ju Zhang, Bowen Yang, Yi Wang, Ye Yang, Yu Qi, Le Yu, Chenhao Jia, Zhanlong Zhang, Xinyu Chen, Xiaobo Xue, Jianyuan Lu, Shize Zhang, Enge Song, Yang Song, Rong Wen, Biao Lyu, Yang Xu, Shunmin Zhu | 12m |
ZooRoute implements tenant‑transparent, underlay‑agnostic failover by changing the VXLAN outer source port and maintaining per‑destination candidate port sets… | ||
| Remote TCP Connection Offload and Applications | Shuo Li, Steven Chien, Tianyi Gao, Michio Honda | 21m |
XO enables direct server return by serializing TLS/TCP connection state and restoring it on backends, cutting L7 proxy CPU and bandwidth. It installs… | ||
| OSCAR: O(1)-Step Convergence and Readily-deployable Congestion Control | Zhaochen Zhang, Feiyang Xue, Rui Ning, Keqiang He, Gianni Antichi, Jiaqi Gao, Zhimeng Yin, Kexin Liu, Rui Li, Zhengqi Cui, Zhehao Lin, Peirui Cao, Guihai Chen, Chen Tian | 19m |
Claims O(1)-step convergence using end-to-end signals rather than INT, implements MIMD updates by estimating delay gradient via batch slope fitting with a… | ||
| Bifrost: Alibaba's Next-Generation VPC Network with High-Performance Multipath... | Zihao Fan, Xing Li, Ye Yang, Bo Jiang, Bowen Yang, Yilong Lv, Yuke Hong, Yinian Zhou, Junnan Cai, Jiayue Xu, Yunrui Hu, Zhao Gao, Ke Sun, Yimin Liu, Xiangdong Zhang, Enge Song, Jianyuan Lu, Xiaoqing Sun, Shize Zhang, Haonan Li, Mingxin Li, Changgang Zheng, Yang Song, Jun Liang, Biao Lyu, Rong Wen, Zhigang Zong, Shunmin Zhu | 17m |
Bifrost moves reliable transport out of the VM into the VPC by inserting a new header and running a transport engine beside AVS on the SmartNIC, using… | ||
| Attack of the Bubbles: Straggler-Resilient Pipeline Parallelism for Large Model Training | Tianyuan Wu, Lunxi Cao, Hanfeng Lu, Xiaoxiao Jiang, Yinghao Yu, Siran Yang, Guodong Yang, Jiamang Wang, Lin Qu, Liping Zhang, Wei Wang | 17m |
PIPEMORPH is a system that adapts pipeline schedules and offloads communication to host memory with CPU-side RDMA delegation to eliminate head-of-line blocking… | ||
| R-TCP: A Framework to Optimize TCP Performance Over Rate-Limiting Networks | Shengtong Zhu, Yan Liu, Lingfeng Guo, Jack Yiu Bun Lee | 16m |
Implements in-kernel Linux changes that detect token-bucket rate limiting by estimating bucket size B and rate R, trigger only when loss exceeds ~20%, then… | ||
| Slowpoke: End-to-end Throughput Optimization Modeling for Microservice Applications | Yizheng Xie, Di Jin, Oğuzhan Çölkesen, Vasiliki Kalavri, John Liagouris, Nikos Vasilakis | 14m |
Slowpoke predicts how a hypothetical optimization changes end-to-end throughput for microservice apps without tracing or a priori call graphs, by slowing down… | ||
| CrossCheck: Input Validation for WAN Control Systems | Alexander Krentsel, Rishabh Iyer, Isaac Keslassy, Bharath Modhipalli, Sylvia Ratnasamy, Anees Shaikh, Rob Shakir | 17m |
CrossCheck validates SDN controller inputs by repairing telemetry via democratic trust propagation and then cross-checking controller topology and demand… | ||
| Detecting and Diagnosing Errors in Serving Archived Web Pages | Jingyuan Zhu, Huanchen Sun, Harsha V. Madhyastha | 16m |
FidEx compares script-level execution traces between crawl and replay and matches pages using layout-tree + JS-write matching, not screenshots, to detect… | ||
| Skyline: A Cloud Centric Internet Monitoring Engine | Shixian Guo, Ziqian Liu, Yangyang Bai, Yuan Chen, Kefei Liu, Qi Zhang, Songlin Liu, Yang Lv, Jianwei Hu, Gen Li, Zhenyang Zhong, Sisi Wen, Yongbin Dong, Feng Luo, Anjian Chen, Rui Han, Jiale Feng, Lingpei Meng, Siwan Chen, Hang Li, Shuai Xu, Juntao Zhong, Chaoran Hu, Yibo Huang, Yiming Qiu | 14m |
Cloud deployment monitors public‑Internet faults with three concrete tools: BiProp uses P4 internet gateways to reply to inbound probes and achieve agentless… | ||
| Feedback-guided Adaptive Testing of Distributed Systems Designs | Ao Li, Ankush Desai, Rohan Padhye | 14m |
Fest uses feedback-guided adaptive schedule generation for formal distributed-system models, implemented in the P framework and evaluated on 94 models. It… | ||
| From Intention to Practice: Towards Systematic Validation of NIDS Rule Enforcement | Huan Liu, Haoyu Chen, Biang Xu, Jingyao Zhou, Bin Yuan, Qiankun Zhang, Deqing Zou, Hai Jin | 20m |
NIDSFuzz reads deployed rules and converts signatures into buffer constraints to synthesize packets that guarantee semantic alignment with targets, then… | ||
| Count-Based Abstractions for Performance Verification of Contention Points | Amir Seyhani, Aarti Gupta, David Walker, Mina Tahmasbi Arashloo | 17m |
Develops a count+window abstraction that tracks small windows of per-class packet counts plus overflow barriers to bound reordering and trade precision for… | ||
| Iceberg: Automated Verification of DNS Authoritative Engines via Just-in-Time... | Yuxing Xiang, Rilin Huang, Naiqian Zheng, Xin Jin | 16m |
Iceberg implements just-in-time summarization over LLVM IR with Z3 to verify DNS authoritative engines automatically, using state splitting, symbolic regions… | ||
| Eywa: Automating Model-Based Testing using LLMs | Rajdeep Mondal, Rathin Singha, Todd Millstein, George Varghese, Ryan Beckett, Siva Kesava Reddy Kakarla | 16m |
Eywa uses LLMs to synthesize modular, executable protocol models (converted to code and composed via dependency graphs) and then drives KLEE-style symbolic… | ||
| Making Logic a First-Class Citizen in Generative ML for Networking | Hongyu Hè, Minhao Jin, Maria Apostolaki | 18m |
NetNomos learns first-order logic rules from four real-world datasets, filters them, then enforces them at inference by running a GPT-2 model alongside an SMT… | ||
| JITServe: SLO-aware LLM Serving with Imprecise Request Information | Wei Zhang, Zhiyu Wu, Yi Mu, Rui Ning, Banruo Liu, Nikhil Sarda, Myungjin Lee, Fan Lai | 14m |
JITServe is an LLM-serving system that schedules with noisy length and dependency estimates, refines them online with quantile regression forests and… | ||
| RollPacker: Taming Long-Tail Rollouts for RL Post-Training with Tail Batching | Wei Gao, Yuheng Zhao, Dakai An, Tianyuan Wu, Lunxi Cao, Shaopan Xiong, Ju Huang, Weixun Wang, Siran Yang, Wenbo Su, Jiamang Wang, Lin Qu, Bo Zheng, Wei Wang | 17m |
RollPacker implements tail batching: it groups prompts that produce very long responses into occasional long rounds and uses speculative execution to run extra… | ||
| FENIX: Enabling In-Network DNN Inference with FPGA-Enhanced Programmable Switches | Xiangyu Gao, Tong Li, Yinchao Zhang, Ziqiang Wang, Xiangsheng Zeng, Su Yao, Ke Xu | 13m |
FENIX implements a hybrid pipeline: a switch-based Data Engine extracts features and uses a probabilistic token bucket to limit feature stream rate, and an… | ||
| DDoS Detection at the Scale of One Hundred Tbps | Yunming Xiao, Xijun Luo, Youliang Jiang, Aike Wang, Hu Chen, Zhibin Zhou, Heng Yu, Jiahao Cao, Yong Jiang, Jilong Wang, Mingwei Xu, Yan Chen, Congcong Miao | 19m |
Canopy implements a distributed DDoS detector that scales past 100 Tbps by combining hash for mice, spray for elephants dynamic load balancing with selective… | ||
| SketchPipe: Toward Accurate Sketch-based Network Measurement on Multi-Pipeline... | Xiang Chen, Longlong Zhu, Linying Zheng, Hongyang Du, Dong Zhang, Jianshan Zhang, Xuan Liu, Qun Huang, Dusit Niyato, Haifeng Zhou, Chunming Wu, Hongyan Liu, Kui Ren | 15m |
Places each sketch array exclusively on one pipeline with splitless sketch placement and asynchronously ships cached flow keys via generated state packets so… | ||
| MORP4: A Dynamic Network Telescope | Iliana Xygkou, Jithin K. Sojan, Dhruv Rauthan, Feng Zhu, Thomas Holterbach, Shane Alcock, Brian Flanagan, Ahmed Saeed, Alberto Dainotti | 18m |
MORP4 implements a programmable data plane telescope in P4 implementation with a Python/C++ controller and runs on a Tofino switch, detecting unused IPv4 at… | ||
| BayWatch: Practical Internet-Scale Topology Monitoring with Dynamic Bayesian Estimation | Zhongxu Guan, Shuai Wang, Li Chen, Zhaoteng Yan, Jiaye Lin, Dan Li, Yong Jiang, Yingxin Wang, Ziqian Liu | 18m |
System that builds a layered model of TTL hops and uses a Dynamic Bayesian Network to infer actual load‑balancing reply (LBR) distributions rather than… | ||
| SyncWise: Error-Aware Time Synchronization for Reconfigurable Data Center Networks | Yiming Lei, Jialong Li, Zhengqing Liu, Raj Joshi, Yiting Xia | 14m |
SyncWise is a synchronization protocol for reconfigurable data center networks that predicts per-node error bounds from preprofiled drift and hop-error limits… | ||
| Building A CSFQ-Inspired Transport for Switched CXL Memory Pooling | Zerui Guo, Emily Shriver, Ming Liu | 16m |
Characterizes an XConn Apollo/Titan switched CXL testbed and shows unmanaged streams cause severe interference — victim bandwidth falls 83% at the host and… | ||
| OpenOptics: Enabling Open Research and Implementation of Optical Data Center Networks | Yiming Lei, Federico De Marchi, Jialong Li, Raj Joshi, Shu-Ting Wang, Xiaoqi Chen, Balakrishnan Chandrasekaran, Yiting Xia | 16m |
OpenOptics provides a concrete software stack and testbed that decouples optical hardware from control: a time-flow table abstraction plus a re‑architected… | ||
| PlanB: Efficient Software IPv6 Lookup with Linearized B+-Tree | Zhihao Zhang, Lanzheng Liu, Chen Chen, Huiba Li, Jiwu Shu, Windsor Hsu, Yiming Zhang | 17m |
PlanB turns IPv6 LPM into a one-dimensional predecessor search over elementary intervals and stores them in a linearized B+-tree laid out for cache lines. It… | ||
| FLARE: Anomaly Diagnostics for Divergent LLM Training in GPU Clusters of Thousand-Plus... | Weihao Cui, Ji Zhang, Han Zhao, Chao Liu, Jian Sha, Bo Sang, Bingsheng He, Minyi Guo, Quan Chen | 17m |
FLARE integrates a lightweight tracing daemon that instruments Python APIs and GPU kernels without backend patches, and a diagnostic engine that performs… | ||
| TurboTest: Learning When Less is Enough through Early Termination of Internet Speed Tests | Haarika Manda, Manshi Sagar, Kartikay Singh, Cindy Zhao, Tarun Mangla, Phillipa Gill, Elizabeth Belding, Arpit Gupta | 14m |
A deployable two-stage system decouples Stage 1 regressor (predicts final throughput from partial TCP info: RTT, retransmissions, congestion window, throughput… | ||
| Defeating Slow-and-Low Threats via Diffusion Model-based Generative Inference | Seyed Mohammad Mehdi Mirnajafizadeh, Prashant Khanduri, DaeHun Nyang, Rhongho Jang | 17m |
SketchVision encodes packet-level timing into compact images in an eBPF-enabled data plane, then uses a diffusion-based denoiser plus a learned forecaster and… | ||
| SwiftEP: Accelerating MoE Inference with Buffer Fusion and TMA Offloading | Xingyi Li, Yadong Liu, Xiaojie Huang, Yiran Zhang, Shuai Wang, Shangguang Wang, Zhehao Lin, Yinben Xia, Chang Yu, Qihang Liu, Xuan Zhang, Hao Lu, Xiang Li, Zekun He, Yachen Wang, Xianneng Zou | 15m |
SwiftEP is an all-to-all library that fuses compute and RDMA buffers and offloads NVLink transfers to the Hopper TMA to remove redundant copies and free SMs… | ||
| Geminet: Learning the Duality-based Topology-Agnostic Update Operator for Lightweight... | Ximeng Liu, Zhuoran Liu, Yingming Mao, Yatao Li, Shizhen Zhao, Xinbing Wang | 18m |
Geminet learns a lightweight, topology-agnostic update operator that iteratively updates edge-level dual variables (congestion prices) and recovers per-pair… | ||
| EROICA: Online Performance Troubleshooting for Large-scale Model Training | Yu Guan, Zhiyu Yin, Haoyu Chen, Sheng Cheng, Chaojie Yang, Kun Qian, Tianyin Xu, Pengcheng Zhang, Yang Zhang, Hanyu Zhao, Yong Li, Dennis Cai, Ennan Zhai | 17m |
EROICA is an online profiler that triggers only on iteration slowdowns, profiles all workers simultaneously, extracts beta, mu, sigma fingerprints for… | ||
| Supercharging Packet-level Network Simulation of Large Model Training via Memoization... | Fei Long, Kaihui Gao, Li Chen, Dan Li, Yiwei Zhang, Fei Gui, Yitao Xing, Wenjia Wei, Bingyang Liu | 17m |
Wormhole implements a C++ ns-3 kernel that uses network partitioning, state memoization and rate-based steady-state detection to skip repetitive packet events… | ||
| Iris: Expressive Traffic Analysis for the Modern Internet | Thea Rossman, Diana Qing, Gerry Wan, Zakir Durumeric | 17m |
Iris implements a compiler that compiles Rust subscriptions into an optimized pipeline by decomposing connection-layer state machines into incremental filters… | ||
| Who Watches the Watchers? On the Reliability of Softwarizing Cloud Application Management | Jiawei Tyler Gu, Zhen Tang, Yiming Su, Bogdan Alexandru Stoica, Xudong Sun, William X. Zheng, Yue Zhang, Akond Rahman, Chen Wang, Tianyin Xu | 16m |
Study analyzes 412 real-world failures across 13 Kubernetes operators and finds 42% stem from operator–application interactions, with 64% of those being… | ||
| From Source to Solution: Tackling Packet Losses in Large-scale Cloud Gaming... | Jing Wang, Xiao Kong, Yunzhe Ni, Nian Wen, Jiaxing Zhang, Congcong Miao, Honghao Liu | 16m |
Production measurements on Tencent START (≈200,000 hours) show 66.5% of losses arise from network bias against UDP and congestion; LADR implements a TCP-first… | ||
| PD3: Prefetching Data with DPUs for Disaggregated Memory | Sidharth Sankhe, Felix Zhang, Umayrah Chonee, Sherman Lim, Jiasheng Hu, JIALIN LI, Qizhen Zhang | 14m |
PD3 implements a DPU-side prefetch path that parses client requests, looks up a bucketized host view, issues RDMA pulls to remote memory, and DMA-loads items… | ||
| UpFuzz: Detecting Data Format Incompatibility Bugs during Distributed Storage System... | Ke Han, Sruthi P C, Yayu Wang, Yaoxu Song, Bishal Basak Papan, Junwen Yang, Pedro Fonseca, Yongle Zhang | 13m |
UPFUZZ builds an automated upgrade fuzzer plus a likely-invariant analysis engine that instruments sync I/O and does backward analysis to capture transitively… | ||
| Libra: Flexible Request Partitioning and Scheduling for Serving Unbalanced and Dynamic... | Chaoyi Ruan, Yinhe Chen, Dongqi Tian, Yandong Shi, Yongji Wu, JIALIN LI, Cheng Li | 17m |
Libra implements a micro-request flexible partitioning that splits requests at any token boundary, a global scheduler that binary-searches a split point (≤6… | ||
| Come Hell or Still Water: Alleviating Tail Latency in Cloud Block Store | Chaolei Hu, Kun Qian, Erci Xu, Yifan Shen, Haoran Zhang, Xue Li, Yuesheng Gu, Lingjun Zhu, Fengyuan Ren, Ennan Zhai | 19m |
Per-proxy dual-bucket token buckets add a main bucket (refilled every 10 ms) and a check bucket that only steady segments can draw from; hot segments drain the… | ||
| Wallet: Confidential Serverless Computing | Patrick Sabanic, Masanori Misono, Teofil Bodea, Julian Pritzi, Michael Hackl, Dimitrios Stavrakakis, Pramod Bhatotia | 18m |
Wallet implements nested confidential execution inside CVMs with a tiny trusted monitor, a decoupled guest OS, LibOS-based zygote images and fork-style… | ||
| Octopus: Enhancing CXL Memory Pods via Sparse Topology | Yuhong Zhong, Fiodar Kazhamiaka, Pantea Zardoshti, Shuwei Teng, Rodrigo Fonseca, Mark D. Hill, Daniel S. Berger | 18m |
Octopus constructs a sparse CXL topology that groups servers into islands (BIBD intra-island, expander inter-island) to trade overlap for pooling efficiency. A… | ||
| RLBoost: Harvesting Preemptible Cloud Resources for Cost-Efficient Reinforcement Learning | Yongji Wu, Xueshen Liu, Haizhong Zheng, Juncheng Gu, Beidi Chen, Z. Morley Mao, Arvind Krishnamurthy, Ion Stoica | 15m |
RLBoost harvests preemptible GPUs (cloud spot instances and spare cluster capacity) to run rollouts while keeping training on reserved full‑mesh GPUs, claiming… | ||
| KUBEDIRECT: Unleashing the Full Power of the Cluster Manager for Serverless Computing | Sheng Qi, Zhiquan Zhang, Xuanzhe Liu, Xin Jin | 16m |
KUBEDIRECT bypasses the API server and uses direct P2P message passing along a sequential controller chain, implementing pairwise agreements as a hierarchical… | ||
| FlexLLM: Token-Level Co-Serving of LLM Inference and Finetuning with SLO Guarantees | Gabriele Oliaro, Xupeng Miao, Xinhao Cheng, Vineeth Kada, Mengdi Wu, Ruohan Gao, Yingyi Huang, Remi Delacourt, April Yang, Yingcheng Wang, Colin Unger, Zhihao Jia | 20m |
FlexLLM co-serves LLM inference and PEFT finetuning on the same GPUs by fusing inference and adapter forward passes and splitting backward work across steps… | ||
| Express Lane to Efficiency and Reliability: Multi-Dimensional Control in Meta’s Express.. | Faisal Iqbal, Vitaly Neganov, Brian Chang, Rong Rong, Yuanjun Yao, Marek Denis, Qin Zhang, Alexandru Manea, Anton Marchenko, Ulas Kozat, Aditya Akella, Ying Zhang | 16m |
Meta’s Express Backbone is eight parallel MPLS/segment‑routing planes and grows with AI traffic; the team adds eight planes aware load balancing via Unequal… | ||
| Sparse Checkpointing for Fast and Reliable MoE Training | Swapnil Gandhi, Christos Kozyrakis | 18m |
MoEvement is an in-memory checkpointing system for MoE models that implements sparse checkpointing, sparse-to-dense conversion, and upstream logging to avoid… | ||
| MAE: More Adaptive Video Encoder for Consistent Low Latency in High-Quality Real-Time... | Hua Meng, Yufan Zhuang, Yasna Noushirvani, Xiangjie Huang, Zili Meng | 16m |
Claims a practical encoder-side fix: MAE tunes encoder internal parameters with a dynamic virtual buffer size and probes congestion-control internals to… | ||
| In Link We Trust: BFT at the Speed of CFT using Switches | Lior Zeno, Naama Ben-David, Mark Silberstein | 17m |
SwitchBFT implements a BFT consensus sequencer on P4 programmable switches, claiming BFT latency and throughput equal to CFT NOPaxos by using packet source… | ||
| Controlling Arbitrary Internet Queues with Titrate | Anchengcheng Zhou, Joshua Lau, P. Brighten Godfrey, Maria Apostolaki | 17m |
Titrate is a router-side closed loop that probes and adapts per-queue thresholds in real time using NS3 packet‑level simulation and live Internet experiments… | ||
| Managing Congestion Control Heterogeneity on the Internet with Approximate Performance... | Ayush Mishra, Archit Bhatnagar, Yixuan Zhang, Ben Leong, Ya Gao, Raj Joshi | 16m |
Claims a practical multi-queue AQM called Santa that groups flows by inferred throughput–delay preference using packet sojourn time and shuffles them every… | ||
| The GOODPUT System: A Machine Learning-Driven Optimization Framework for Dynamic... | Yu Wang, Robert P. Dick | 11m |
Optimizes application-layer goodput by jointly choosing AP channel assignments, channel bonding decisions, and station→AP mappings using LSTM demand… | ||
| CStar Gateway: Augmenting Public Cloud Infrastructure for Heterogeneous Network... | Haonan Li, Tian Pan, Jin Ke, Baohai Hu, Changgang Zheng, Enge Song, Zhi Xu, Ye Yang, Bowen Yang, Donglin Lai, Yisong Qiao, Bengbeng Xue, Jianyuan Lu, Xiaoqing Sun, Shize Zhang, Zihao Fan, Mingxin Li, Yang Song, Jun Liang, Xionglie Wei, Biao Lyu, Rong Wen, Zhigang Zong, Jiao Zhang, Tao Huang, Shunmin Zhu | 18m |
CStar Gateway shifts NF VM multi-tenancy from the vSwitch into each NF VM using a Tofino-based stateless gateway and GENEVE+VXLAN nested tunneling so each NF… | ||
| DistRS: Disaggregated Reward Service for RLVR with Batch-Level Constraint | Ruidong Zhu, Mingcong Han, Yinmin Zhong, Wencong Xiao, Xuanzhe Liu, Xin Jin | 16m |
DistRS is a disaggregated reward service that separates reward computation from training, implements a microservice pipeline with per-stage worker pools, and… | ||
| MuxTune: Efficient Multi-Task LLM Fine-Tuning in Multi-Tenant Datacenters... | Chunyu Xue, Yi Pan, Weihao Cui, Quan Chen, Shulai Zhang, Bingsheng He, Minyi Guo | 15m |
Datacenter fine-tuning system that multiplexes a frozen LLM backbone across concurrent PEFT tasks, using spatial-temporal multiplexing and a hierarchical… | ||
| CCC: Re-architecting Delay-based Congestion Control in Datacenter Networks | Wanchun Jiang, Haoyang Li, Kai Wang, Yujie Hu, Xiao Han, Danfeng Shan, Fengyuan Ren, Jiawei Huang, Jianxin Wang | 18m |
Claims that delay signals alone can match switch-assisted CC by adding a CiD trustworthiness layer that splits operating space into regions and switches… | ||
| Starfish: A Topology-Routing Co-Design for Small-Scale Data Centers | Anchengcheng Zhou, Vipul Harsh, Sangeetha Abdu Jyothi, Maria Apostolaki, Brighten Godfrey | 19m |
Starfish rearranges existing switches into a DRing double‑ring topology and implements a Shortest Near K routing scheme using VRF partitions and BGP (prototype… | ||
| HEDGE: Traffic Engineering with Probabilistic Link Capacities | Arjun Devraj, Bill Owens, Umesh Krishnaswamy, Ying Zhang, Rachee Singh | 17m |
Lab experiments and a production cloud trace show that individual wavelengths on the same fiber fail independently enough to make link capacity a discrete… | ||
| CascadeNet: Generating Network Traffic with High-Fidelity Temporal Patterns | Runwei Lu, Yanran Deng, Ruixuan Li, Jinting Liu, Yuejie Wang, Xinyu Li, Deming Xu, Han Tian, Kai Chen, Guyue Liu | 15m |
CascadeNet generates traces by converting each flow into an aggregated per-flow time series and then using CascadeGAN hierarchical generators (MLP → LSTM →… | ||
| A Composable Emulation Framework for Whitebox Switches | Congcong Miao, Xianneng Zou, Chuwen Zhang, Shiping Yang, Qihang Liu, Zhijie Yan, Yanke Zhang, Yong Jiang, Qiao Xiang, Xin Jin, Zili Meng, Ang Chen | 20m |
MirSwitch composes a production SONiC control plane with a modified VPP data plane and virtualized peripheral modules using an adapter-centric design: EVPP… | ||
| Harp: Improving VPC Network Availability via Efficient Failure Detection and Rerouting... | Jiayu Hu, Feng Jin, Xianping Zhou, Kai Zhang, Zhen Shen, Yongkang Luo | 24m |
Software system that monitors feasible physical underlay paths per VM pair and embeds inband inspection probes into normal packets, then flips the underlay UDP… | ||
| Themis: Detecting Distributed Concurrency Bugs through RPC-Driven Race-Directed Test... | Hongchen Cao, Jingzhu He, Ting Dai, Guoliang Jin | 16m |
Themis combines static RPC-driven race detection with LLM-based test generation and directed fuzzing and interleaving exploration to find distributed… | ||
| Bridging Storage and Execution: A Semantic Virtual Bus for On-Demand Application Streaming | Jun Lu, JIALIN LI, Yaoxue Zhang, Ju Ren | 16m |
STREAMBUS implements a semantic virtual storage bus that turns remote storage into an execution-aware layer by combining a dual-mode transmission (fast… | ||
| QCON: Seamless QoE-Aware 5G Streaming via Multi-Connectivity | Goodsol Lee, Junhong Min, Seyeon Kim, Juheon Yi, Kwang Taik Kim, Mung Chiang, Sangtae Ha, Kyunghan Lee, Saewoong Bahk | 16m |
QCON leverages 5G multi‑connectivity inside the RAN to infer frame‑level QoE by parsing RTP headers and to drive a QoE Monitor that only uses backup 4G… | ||
| Law: Towards Consistent Low Latency in 802.11 Home Networks | Yibin Shen, Zili Meng | 15m |
Law restructures the 802.11 link layer to enforce a latency bound by unified queuing, per-transmission-opportunity rate updates, active early drops, and a… | ||
| Enabling SLO-Aware 5G Multi-Access Edge Computing with SMEC | Xiao Zhang, Daehyeok Kim | 18m |
SMEC implements decoupled RAN and edge schedulers that infer deadlines from existing signals—5G control signals at the RAN and a lightweight probe-and-ack… | ||
| Learning to Tune Optical WANs: A Field Deployment of Noise Models in Optical Networks | Bhaskar Kataria, Howard Hua, Andrea D Amico, Bill Owens, Rachee Singh | 16m |
Builds a Gaussian Noise model of a New York ISP optical backbone including amplifiers, fiber spans, ROADMs and transceivers and achieves under 4% modeling… | ||
| Enabling AI Network Cross-Layer Design and Operations with Arcadia... | Zhaodong Wang, Satyajeet Singh Ahuja, Xu Zhang, Yuhui Zhang, Max Noormohammadpour, Gregory R. Steinbrecher, Thomas Fuller, Xin Liu, Kevin Quirk, Mikel Jimenez Fernandez, Abhinav Triguna, Yan Cai, Steve Politis, Petr Lapukhov, Naader Hasani, Ying Zhang | 19m |
Arcadia is a calibrated, multi‑fidelity discrete‑event simulator that replays production traces to answer counterfactuals about schedulers, transports… | ||
| Phantora: Maximizing Code Reuse in Simulation-based Machine Learning... | Jianxing Qin, Jingrong Chen, Xinhao Kong, Yongji Wu, Tianjun Yuan, Liang Luo, Zhaodong Wang, Ying Zhang, Tingjun Chen, Alvin R. Lebeck, Danyang Zhuo | 17m |
Phantora runs unmodified ML frameworks inside containerized rank emulators and intercepts CUDA/NCCL calls to forward computation and communication events to a… | ||
| PrvTel: Lightweight Models for Private and Accurate Telemetry Data Retention | Yajie Zhou, Fuheng Zhao, Eric Wang, Ayse K. Coskun, Divyakant Agrawal, Amr El Abbadi, Zaoxing Liu | 16m |
PrvTel builds and ships a compact generative retention system that trains a domain-specialized variational autoencoder on privacy-noised telemetry and keeps… | ||
| ServeGen: Workload Characterization and Generation of Large Language... | Yuxing Xiang, Xue Li, Kun Qian, Yan Zhang, Wenyuan Yu, Ennan Zhai, Xin Jin, Jingren Zhou | 16m |
Characterization uses production traces spanning 3.5 billion requests over 4 months across 10 models and finds that traffic is dominated by a few customers —… | ||
| REAL: Emulating Control Plane at Simulator’s Cost | Ze Xia, Hao Li, Jinyu Fu, Xin Wan, Yihan Dang, Danfeng Shan, Li Chen, Peng Zhang | 15m |
REAL implements transparent syscall interposition via LD_PRELOAD, replaces TCP over veth with message-level IPC delivery, enforces run-to-idle scheduling and… | ||
| ZOC: Elastic and Cost-Efficient Virtual SmartNIC Architecture for Cloud Physical Machines | Naixuan Guan, Xiaokang Hu, Yisheng Xie, Xishi Qiu, Chaojie Liu, Yuchao Cao, Banghao Ying, Dianchen Tian, Yu Zhou, Yangzeyu Zhang, Hujun Ge, Yibin Shen, Jiesheng Wu | 17m |
ZOC implements a software-defined virtual SmartNIC by combining a passthrough NIC with a dedicated service VM that traps MMIO to an IO proxy, uses a virtual… | ||
| SLATE: Service Layer Traffic Engineering | Gangmuk Lim, Aditya S Prerepa, Brighten Godfrey, Radhika Mittal | 18m |
SLATE is a service-layer traffic engineering system that optimizes multi-cluster microservice request routing and runs as a web-service controller plus an… | ||
| A Fast Solver-Free Algorithm for Traffic Engineering in Large-Scale Data Center Network | Yingming Mao, Qiaozhu Zhai, Ximeng Liu, Zhen Yao, Xia Zhu, Yuzhou Zhou | 13m |
Sequential Source-Destination Optimization (SSDO) decomposes DCN TE into per source‑destination subproblems and replaces per‑subproblem LP solves with a… | ||
| AVA: Towards Agentic Video Analytics with Vision Language Models | Yuxuan Yan, Shiqi Jiang, Ting Cao, Yifan Yang, Qianqian Yang, Yuanchao Shu, Yuqing Yang, Lili Qiu | 23m |
AVA builds an Event Knowledge Graphs index from long streams by windowed captioning with a 7B VLM (e.g., ChainOne VL), semantic merging via BLEU/BERT-like… | ||
| Themis: Scheduling-Aware Buffer Management for HBM-Based Hybrid Buffers | Zhiyu Zhang, Minkun Xue, Kan Yu, Ruyi Yao, Shili Chen, Hao Mei, Zixuan Chen, Weiyi Chen, Yibo Fan, Yang Xu | 17m |
Themis places packets by P4 rank, keeps rank-based admission to favor soon-to-depart packets in SRAM, and uses batch swap in/out to write and read contiguous… | ||
| SmartNIC-Enabled Live Migration for Storage-Optimized VMs with PYROCUMULUS | Jiechen Zhao, Ran Shu, Lei Qu, Ziyue Yang, Rui Ma, Derek Chiou, Natalie Enright Jerger, Peng Cheng, Yongqiang Xiong | 17m |
Pyrocumulus offloads live migration of storage-optimized VMs into an FPGA SmartNIC running ~16K lines of SystemVerilog on an Intel Arria 10, attached to Intel… | ||
| Uber's Failover Architecture: Reconciling Reliability and Efficiency in Hyperscale... | Mayank Bansal, Milind Chabbi, Kenneth Sejdenfaden Bøgh, Srikanth Prodduturi, Kevin Xu, Amit Kumar, David Bell, Ranjib Dey, Yufei Ren, Sachin Sharma, Juan Marcano, Shriniket Kale, Subhav Pradhan, Ivan Beschastnikh, Miguel Covarrubias, Chien-Chih Liao, Sandeep Koushik Sheshadri, Wen Luo, Kai Song, Ashish Samant, Sahil Rihan, Nimish Sheth, Albert Greenberg, Uday Kiran Medisetty | 20m |
Claims a shift from uniform 2× regional provisioning to a differentiated model that cuts steady‑state from 2× to 1.3×, raises utilization from ~20% to ~30%… | ||
| SYMPHONY: Enabling Compute-Memory Disaggregation in LLM Serving Systems | Saurabh AGARWAL, Bodun Hu, Anyong Mao, Aditya Akella, Shivaram Venkataraman | 14m |
SYMPHONY builds a low-latency disaggregated memory layer that moves KV caches off GPUs using advisory requests prefetch and a multi-tier storage pipeline… | ||
| Defending against Traffic Analysis Attacks with Flexible In-Network Obfuscation | Guorui Xie, Qing Li, Zhenning Shi, Gianni Antichi, Yijia Zhu, Kejun Li, Changxing Weng, Sebastiano Miano, Yong Jiang, Mingwei Xu | 19m |
Securitas is an in-network obfuscation framework that uses packet fragmentation and small-TTL packet insertion (triggering ICMP replies) and trains an LSTM… | ||
| ForestColl: Throughput-Optimal Collective Communications on Heterogeneous Network Fabrics | Liangyu Zhao, Saeed Maleki, Yuanhong Wang, Zezhou Wang, Ziyue Yang, Hossein Pourreza, Arvind Krishnamurthy | 15m |
ForestColl generates throughput-optimal collective schedules by computing the bottleneck network cut via an optimality binary search, transforming switch… | ||
| Matryoshka: Realizing Hyperscale Data Center Network Design for the AI Era | Yan Cai, Jialong Li, Kutalmis Akpinar, Tianxiang Li, Hany Morsy, Jason Wilson, Sunil Khaunte, Yiting Xia, Ying Zhang | 16m |
Matryoshka compiles high-level Thrift network intents into vendor-specific switch configs, modeling topology, hierarchical IP allocation, BGP and routing… | ||
| PlanetServe: A Decentralized, Scalable, and Privacy-Preserving Overlay for... | Fei Fang, Yifan Hua, Shengze Wang, Ruilin Zhou, Yi Liu, Chen Qian, Xiaoxue Zhang | 16m |
PlanetServe builds a peer-to-peer overlay that uses Theta secure information dispersal (Rabin IDA + Shamir shares) for anonymous prompts, an HR tree hash… | ||
| MoCE: A Mixture-of-Context Aware Experts Framework for Troubleshooting... | Vipul Harsh, Sayan Sinha, Henry Milner, B. Aditya Prakash, Vyas Sekar, Hui Zhang | 18m |
MoCE implements a mixture-of-context-aware experts system where each expert is expressed as a dataflow DAG of SQL, stateful, and statistical operators; a… | ||
| Heuristic Analysis from Source Code via Symbolic-Guided Optimization | Pantea Karimi, Siva Kesava Reddy Kakarla, Ryan Beckett, Santiago Segarra, Pooria Namyar, Mohammad Alizadeh, Behnaz Arzani | 15m |
MetaEase analyzes heuristic source code directly, using symbolic execution (KLEE) to generate seeds per code path and a duality-based gradient for the optimal… | ||
| CCEval: Accurately and Confidently Evaluating Performance Metrics of Congestion... | Tianfeng Liu, Kaihui Gao, Li Chen, Dan Li, Jin Guang, Xinyun Chen, Vincent Liu, Zhiyong Chen, Yiwei Zhang, Ni Jin, Ran Zhang | 17m |
CCEval is an evaluation framework that outputs confidence intervals and forecasts needed trials using model-free trial forecasting and speeds extreme-tail… | ||
| Predict, Prune, Play: Efficient Video Playback Optimization Under Device... | Harsha Sharma, Pouya Hamadanian, Arash Nasr-Esfahany, Zahaib Akhtar, Mohammad Alizadeh | 17m |
Large RCT of 750 million sessions across 84 production-grade treatments and 53 device models shows best playback settings vary by device and cohort, so AZEEM… | ||
| Over-Threshold Multiparty Private Set Intersection for Collaborative... | Onur Eren Arpaci, Raouf Boutaba, Florian Kerschbaum | 17m |
Protocol identifies IPs seen by at least t=3 parties using Shamir-style secret shares and a novel binning/hashing scheme that uses 20 hash tables sized… | ||
| Secure Vickrey Auctions for Online Advertising | Archit Bhatnagar, Yunming Xiao, Ang Chen, Amrita Roy Chowdhury | 17m |
Obsidian implements a maliciously secure Vickrey auction using a new reverse unary encoding for bids, verifiable function secret sharing to push validation… | ||
| EZ-SAVE: Evaluation of Easy-to-Deploy Source Address Validation Policies | Nicholas Scaglione, Justin Furuness, Yossi Gilad, Hemi Leibowitz, Cameron Morris, Bing Wang, Kotikalapudi Sriram, Amir Herzberg | 14m |
Evaluates easy-to-deploy SAV policies with a modified BGPI simulator on realistic KADA topology and RIPE/RouteViews measurements, running 1,000 trials per… | ||
| cc-pipe: Breaking Systemic Bottlenecks in RPKI Data Supply Chain with Concurrent and... | Chenhui Yu, Yanbiao Li, Hui Zou, Yuxuan Chen, Shiyi Liu, Gaogang Xie | 18m |
cc-pipe replaces the RP monolithic validation barrier with a concurrent pipeline that generates local view updates per resource certificate, uses a predictive… | ||
| A Systematic Threat Analysis and Practical Attacks on Automated Frequency Coordination | Yilu Dong, Tianchang Yang, Arupjyoti Bhuyan, Syed Rafiul Hussain | 17m |
AFC systems let unauthenticated inputs (GPS, Wi‑Fi positioning, DNS, NTP) control 6 GHz grants and are practically exploitable: an off‑path attacker using a… | ||
| Mortise: Auto-tuning Congestion Control to Optimize QoE via Network-Aware Parameter | Yixin Shen, Ruihua Chen, Bo Wang, Jing Chen, Haochen Zhang, Minhu Wang, Yan Liu, Mingwei Xu, Zili Meng | 15m |
Mortise auto-tunes rule‑based congestion control in real time by inferring an application's throughput‑delay‑loss preference and mapping that to CCA… | ||
| Harvesting Spare CPU Resources in Container Systems | Adam Hall, Anirudh Sarma, Esha Choukse, Umakishore Ramachandran, Sameh Elnikety | 13m |
HarvestContainers is a Linux kernel module plus user daemon integrated with Kubernetes that dynamically reclaims ephemeral cores from overprovisioned… | ||
| Fractal: Fault-Tolerant Shell-Script Distribution | Zhicheng Huang, Ramiz Dundar, Yizheng Xie, Konstantinos Kallas, Nikos Vasilakis | 13m |
FRACTAL implements fault-tolerant distribution for unmodified shell scripts by treating each worker-assigned fragment as an opaque subgraph and tracking only… | ||
| eXpressSFU: Toward Super-Scalable Video Conferencing with SmartNICs | Tuan Tran, S. M. H. Hosseini, Seyeon Kim, Kyunghan Lee, Nam Bui, Dirk Grunwald, Sangtae Ha | 15m |
eXpressSFU re-architects an SFU by decoupling fast-control and data planes and offloading packet operations to a commodity BlueField-2 SmartNIC, implementing a… | ||
| ZipLLM: Efficient LLM Storage via Model-Aware Synergistic Data Deduplication and | Zirui Wang, Tingfeng Lan, Zhaoyuan Su, Juncheng Yang, Yue Cheng | 13m |
Hugging Face model storage grows ~275× from 2023 to 2025 to ~77 PB and fine-tuned variants dominate, so ZipLLM targets cross-model redundancy with tensor-level… | ||
| Latency-Aware Caching with Delayed Hits: From Bursty Traffic to Pipeline Architectures | Nadav Keren, Gil Einziger, Gabriel Scalosub | 16m |
Proposes an adaptive pipeline cache that composes simple eviction blocks (LRU, LFU, etc.) into ordered stages with configurable quanta and a runtime… | ||
| Cortex: Achieving Low-Latency, Cost-Efficient Remote Data Access For LLM via | Chaoyi Ruan, Chao Bi, Kaiwen Zheng, Ziji Shi, Xinyi Wan, JIALIN LI | 17m |
Cortex caches semantically similar LLM queries by storing a Semantic Element (query, response, and metadata: cost, latency, frequency, size, staleness) and… | ||
| Unleashing The Potential of Datacenter SSDs by Taming Performance Variability | Gohar Irfan Chaudhry, Ankit Bhardwaj, Zhenyuan Ruan, Adam Belay | 17m |
Sandook is a rack-scale block storage system with a logically centralized controller plus lightweight clients and SSD agents that implements log-structured… | ||
| Agentix: An Efficient Serving Engine for LLM Agents as General Programs | Michael Luo, Xiaoxiang Shi, Colin Cai, Tianjun Zhang, Justin Wong, Yichuan Wang, Chi Wang, Yanping Huang, Zhifeng Chen, Joseph E. Gonzalez, Ion Stoica | 16m |
Agentix intercepts LLM calls from programs, tags requests with program and thread IDs, and maintains a process table of program-level stats to drive decisions… | ||
| Queue-Mem: Energy-Efficient Hardware Storage for Advanced Network Function Acceleration | Mariano Scazzariello, Tommaso Caiazzi, Hamid Ghasemirahni, Dejan Kostić, Marco Chiesa | 16m |
Queue-Mem stores packet payloads in the switch shared packet buffer by pausing/resuming port queues with a batch-based queue control that maps five-tuple… | ||
| Hierarchical Integration of WebAssembly in Serverless for Efficiency and Interoperability | Mohammadamin Baqershahi, Changyuan Lin, Visal Saosuo, Paul Chen, Mohammad Shahrad | 17m |
A hierarchy places many WebAssembly modules inside larger container sandboxes to preserve container semantics while packing tenants for density; implementation… | ||
| Mitigating CPU Frontend for Complex Data Plane Applications | Yihan Dang, Hao Li, Ze Xia, Jiajun Luan, Peng Zhang | 17m |
Finds CDPAs stall on the CPU front end (45.1% of cycles on average) due to instruction working sets far larger than L1 I‑cache, and builds NanoPL: an offline… | ||
| FAST: An Efficient Scheduler for All-to-All GPU Communication | Yiran Lei, Dongjoo Lee, Liangyu Zhao, Daniar Kurniawan, Chanmyeong Kim, Heetaek Jeong, Changsu Kim, Hyeonseong Choi, Liangcheng Yu, Arvind Krishnamurthy, Justine Sherry, Eriko Nurvitadhi | 18m |
FAST reframes All-to-All(v) scheduling by treating scale-up (NVLink/local GPU) as effectively free and optimizing only scale-out transfers, implementing… | ||
| HeteCCL: Synthesizing Near-Optimal Collective Communication Schedules for Heterogeneous | Chenyang Hei, Fuliang Li, Jiayi Li, Jiamin Cao, Chengxi Gao, Xiuzhu Sha, Tongrui Liu, Dengke Zhang, Ennan Zhai, Xingwei Wang | 14m |
HeteCCL generates near‑optimal collective schedules by quantizing data into step‑level chunks, modeling topology and link bandwidth precisely, and casting… | ||
| Medley: Optimizing Midgress Bandwidth for Commercial Live Streaming CDNs | Haiping Wang, Wanxin Shi, Sandesh Dhawaskar Sathyanarayana, Shu Shi, Feng Qian, Yinghao Yang, La Zuo, Hebin Yu, Ruixiao Zhang, Ruoshi Sun, Yajie Peng, Xiaofei Pang, Ruili Fang, Zhenpeng Zhu, Weiqi Chen, Yang Xu | 16m |
Medley implements a substream-based CDN that splits each live video into k substreams so each edge node caches 1/k of the stream, claiming a 76.83% midgress… | ||
| Offloading Cloud Network Services at Production Scale | Shaofeng Wu, Zhixiong Niu, Riff Jiang, Lawrence Lee, Junhua Zhai, Ze Gan, Vasundhara Volam, Prabhat Aravind, Prince Sunny, Prince George, Qi Luo, Evan Langlais, Soumya Tiwari, Venkat Satish Katta, Weixi Chen, Rishiraj Hazarika, Sachin Jain, Deven Jagasia, Michal Zygmunt, Avijit Gupta, Neeraj Motwani, Pranjal Shrivastava, Qiang Su, Anil Reddy Pannala, Kristina Moore, James Grantham, Anupam Pandey, Xin Liu, Guohan Lu, Gerald De Grace, Rishabh Tewari, Lihua Yuan, Erica Lan, Deepak Bansal, Dave Maltz, Yongqiang Xiong, Hong Xu | 15m |
Microsoft deploys SONiC DASH SmartSwitch: an immutable DASH pipeline programming model plus a uni-box SmartSwitch that co-locates NPU and up to eight DPUs and… | ||
| AnyPro: Preference-Preserving Anycast Optimization based on Strategic AS-Path Prepending | Minyuan Zhou, Yuning Chen, Jiaqi Zheng, Yifei Xu, Pan Hu, Yongping Tang, Wendong Yin, Jie Lin, Qingyan Yu, Yuanchao Su, Guihai Chen, Wanchun Dou, Songwu Lu, Wan Du | 19m |
AnyPro is a system that uses AS-path prepending and an ICMP probing testbed to derive constraint-based ASPP settings that steer clients to preferred PoPs. It… | ||
| HCDN: Coordinated Stream Scheduling for Cost-Effective Live Video Delivery | Liying Wang, Jing Liu, Yuhan Zhou, Chengke Wang, Mingming Lu, Qingyue Li, Song Geng, Linsen Wang, Kaida Hu, Haoyuan Huang, Shimao Tian, Ri Lu, Mingfei Hao, Chenren Xu, Shu Shi | 14m |
HCDN augments a live CDN by adding ISP-edge and multi-homed nodes and a central scheduler called OpenTiga to run redirection strategies: frozen-stream… | ||
| XLL: Cross-Layer Logging for Data Deduplication in Consensus-Based Storage | John Shawger, Arnav Jhingran, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau | 15m |
A shared append-only log called XLL replaces separate consensus and KV write-ahead logs and stores values once while the KV LSM keeps only locators; integrated… | ||
| HyperEdge: An Edge CDN Infrastructure for Cost Efficient Video Streaming | Dehui Wei, Jiao Zhang, Haozhe Li, Rui Han, Zhichen Xue, Yajie Peng, Xiaofei Pang, Yan Ma, JIALIN LI | 19m |
HyperEdge is a hybrid CDN that combines an external CDN for initial segments with a pool of underutilized devices, using multi-path transmission and a… | ||
| Co-Designing Traffic Control with NVMe-oF for Disaggregated Storage: A Comparative Study | Chendong Wang, Joontaek Oh, Ming Liu | 17m |
Compares switched and switchless NVMe-oF SANs using small real prototypes and large simulations, NVMe over TCP, BlueField-2 SmartNICs, and Samsung PM983 SSDs… | ||
| Cost-effective and Reliable Global Internet Peering with Programmable Switches | Congcong Miao, Zhiyi Yao, Jianchao Lv, Jinglin Wang, Shihan Lin, Xinyi Zhang, Yunming Xiao, Wei Guo, Jiwu Bu, Yachen Wang, Xianneng Zou, Yong Jiang, Marco Canini, Gaogang Xie | 15m |
A production hybrid peering system called Janus uses programmable switches (Broadcom Trident3 in testbed) plus DPDK 48-core hosts to offload inbound traffic to… | ||
| KRAKENGUARD: Towards Fine-Grained eBPF Isolation | Jainil Patel, Lucas Graeff Buhl-Nielsen, Adrien Ghosn, Marios Kogias | 17m |
KRAKENGUARD is a privileged user‑space manager that enforces JSON policies at load time by lifting eBPF bytecode to LLVM IR and running exhaustive symbolic… | ||
| OneSidedMW: Managing Disaggregated Memory Efficiently, Flexibly, and Securely with RNIC | Zixuan Wang, Jinyu Gu, Xingda Wei, Yubin Xia | 16m |
System implements RNIC offloading plus memory windows to perform remote bind/unbind on the NIC so allocation and deallocation happen with zero memory-node CPU… | ||
| Syntra: Synthesizing Cross-Layer Controllers for Low-Latency Video Streaming | Jia Pan, Anup Agarwal, Işıl Dillig, Venkat Arun | 16m |
A tool called Syntra synthesizes a cross-layer controller from a symbolic worst-case model and a declarative, lexicographic objective, using bounded-horizon… | ||
| Net-P4ct: Enhanced WAN Bandwidth Fair Sharing Using P4 Programmable Switches | Haoran Chen, Mingwei Cui, Yihan Zou, Yihang Miao, Suhan Jiang, Damu Ding, Lirong Lai, Ming Gao, Rui Jiang, Shengyuan He, Anjian Chen, Jiaming Shi, Junjie Wan, Yandong Duan, Ruomin Fang, Hongyu Wu, Yongping Tang, Qiao Kang, Guangrui Wu, Xiyun Xu | 19m |
Net-P4ct moves QoS control into P4 programmable switches and pairs lightweight NetAgent eBPF tags with a centralized controller that runs weighted max‑min fair… | ||
| FRCC: Towards Provably Fair and Robust Congestion Control | Anup Agarwal, Venkat Arun, Srinivasan Seshan | 16m |
Claims that FRCC provably bounds unfairness under jitter by encoding only link fraction (flow count) into congestion signals and separately probing capacity… | ||
| nothing matches | ||