ACM Symposium on Cloud Computing 2026

Shangri-La Rasa Sentosa Resort, Sentosa, Singapore

November 18-20, 2026

SoCC 2026 Detailed Program

ACM Symposium on Cloud Computing 2026 · November 18–20, 2026 · Shangri-La Rasa Sentosa Resort, Sentosa, Singapore

Parallel sessions are shown side by side. Full and industry papers and sponsored talks have 20-minute slots; short and vision papers have 15-minute slots. See the program at a glance for a compact overview.

Day 1 · Wednesday, November 18, 2026

Time
Track A · Siloso 2
Track B · Siloso 3
09:00 – 09:30
Opening RemarksWelcome by the General and Program Chairs
09:30 – 11:00
Keynote 1: Haibo ChenSiloso 2 + 3
11:00 – 11:45
Coffee Break
11:45 – 12:45
Track A · Siloso 2
Session 1A: LLM Serving: Parallelism & Model Loading
Chair: Peter Pietzuch
  1. 11:45
    Slate: Execution-Aware Checkpointing for Dynamic Pipeline-Parallel LLM Serving
    Zhaorui Wu, Xiaoping Li, Shujie Pang (Guangdong University of Technology), Yuhui Deng (Jinan University), Geyong Min (University of Exeter)
  2. 12:05
    PipeLive: Efficient Live Pipeline Parallelism Reconfiguration for Dynamic LLM Serving
    Xu Bai, Muhammed Tawfiqul Islam (University of Melbourne), Chen Wang (IBM Research), Adel N. Toosi (University of Melbourne)
  3. 12:25
    MG-MoE: Accelerating Multi-GPU MoE Inference via Expert Prefetching and Cache Coordination
    Guangqiang Luan, Pu Pang, Quan Chen, Anbang Wu, Jingwen Leng (Shanghai Jiao Tong University), Minyi Guo (Guizhou Provincial Laboratory of Big Data, Guizhou University & Shanghai Jiao Tong University)
Track B · Siloso 3
Session 1B: Security, Isolation & Trust
Chair: Kenji Kono
  1. 11:45
    The Kernel Always Knows: Understanding Isolation via Kernel Objects
    Anjali, Michael M. Swift (University of Wisconsin-Madison)
  2. 12:05
    ExclaveFL: Verifiable Claims for Federated Learning using Exclaves
    Jinnan Guo (Imperial College London), Kapil Vaswani (affiliation TBD), Andrew Paverd (Microsoft Security Response Center), Peter Pietzuch (Imperial College London)
  3. 12:25
    Sharding Capsule: Verifiable Prediction for Secure Adaptive Blockchain Sharding
    Jinze Du (Tianjin University of Technology), Xiaohan Liu, Leixiao Li, Hao Lin (Inner Mongolia University of Technology; Inner Mongolia Key Laboratory of Beijiang Cyberspace Security), Qingbo Hao (Tianjin University of Technology), Hanyu Feng (Inner Mongolia University of Technology; Inner Mongolia Key Laboratory of Beijiang Cyberspace Security)
12:45 – 14:15
Lunch
14:15 – 15:35
Track A · Siloso 2
Session 1C: GPU Sharing & Memory Management
Chair: TBA
  1. 14:15
    BoxD: Managing GPU Managed Memory
    Utkarsh Tiwari, Purushottam Kulkarni (Indian Institute of Technology Bombay)
  2. 14:35
    SAllocator: Decoupling Memory Management from Execution for Shared-Friendly Multi-Tenant GPU Inference
    Zehua Yang, Zhongwang Fu (Peking University), Yifeng Tang, Yi Ji (Huawei Technologies), Zhenlin Wang (Michigan Tech), Yingwei Luo, Xiaolin Wang, Diyu Zhou (Peking University)
  3. 14:55
    GRIP: In-place Spatial Partitioning of CUDA Graphs for Cloud-Native Inference Serving
    Sangho Yeo, Jiheon Kim, Sejin Oh (Ten Inc.)
  4. 15:15
    Tessera: Contract-Driven Cost–SLO Optimization for Multi-tenant GPU Clouds
    Rui Li, Shuang Cao (Hill Research)
Track B · Siloso 3
Session 1D: Replication & Consensus
Chair: TBA
  1. 14:15
    TRAM: Tree-based atomic multicast on RDMA
    Lorenzo Martignetti, Eliã Batista (Università della Svizzera italiana), Gianpaolo Cugola (Politecnico di Milano), Fernando Pedone (Università della Svizzera italiana)
  2. 14:35
    Robust and efficient replication with CAV data structures
    Eliã Batista (Università della Svizzera italiana), Eduardo Alchieri (University of Brasilia), Paulo Coelho (Federal University of Uberlandia, Brazil), Fernando Dotti (PUC), Fernando Pedone (Università della Svizzera italiana)
  3. 14:55
    MIDAS: Multi-core Scalable Replication for Key-Value Stores via Distributed Linearization
    Dohyun Kim, Sukjoon Oh, Youjip Won (KAIST)
  4. 15:15
    Polaris: A Coordination Protocol for Planet-Scale BFT Consensus
    Shashank Joshi, Qingnan Duan, Linguan Yang, Hanming Lu, Bernard Wong, Wojciech Golab (University of Waterloo), Stephen Lee (University of Pittsburgh), Srinivasan Keshav (University of Cambridge), Prashant Shenoy (University of Massachusetts Amherst), Sergey Gorbunov, Naman Sood (University of Waterloo)
15:35 – 16:20
Coffee Break
16:20 – 17:40
Track A · Siloso 2
Session 1E: ML Training Systems
Chair: TBA
  1. 16:20
    ZeRO-Emu: Backend-Aligned Execution-Driven Emulation for Parameter-Sharded LLM Training
    Siyuan Yang (National University of Defense Technology; Qiyuan Lab), Enda Yu (National University of Defense Technology), Dezun Dong (National University of Defense Technology; Qiyuan Lab), Si Zheng (Qiyuan Lab)
  2. 16:40
    SKAI: Scaling Industry-Scale Recommendation Model Training with Embedding OptimizationIndustry
    Junhong Liu (Beihang University), Jin Ouyang (Unaffiliated), Xiaoyang Sun (Kuaishou Inc.), Jiapeng Chen, Guowang Zhang (Kuaishou Inc.), Yitang Yang, Dinghao Xue, Yuxi Luo, Zeying Wang (Beihang University), Haiping Xu, Hengrui Ju, Xu He, Yiwu Liu, Chen Yang (Kuaishou Inc.), Tianyu Wo, Chunming Hu, Renyu Yang (Beihang University)
  3. 17:00
    xRLHF: Efficient RLHF with Serverless Computing
    Rui Wei, Hanfei Yu (Stevens Institute of Technology), Devesh Tiwari (Northeastern University), Jian Li (Stony Brook University), Seung-Jong Park (Missouri University of Science & Technology), Hao Wang (Stevens Institute of Technology)
  4. 17:20
    Allpass: Adaptive, Cost-Efficient, Serverless Scheduling for LLM Fine-Tuning across Heterogeneous GPU Clouds
    Yuqi Fu (University of Virginia), Ruizhe Shi (George Mason University), Tingfeng Lan, Zhaoyuan Su, Rui Yang (University of Virginia), Haoliang Wang (Adobe Research), Songqing Chen (George Mason University), Yue Cheng (University of Virginia)
Track B · Siloso 3
Session 1F: Operating Systems & Memory
Chair: Fernando Pedone
  1. 16:20
    Not All Huge Pages Are Real: Detecting and Avoiding Misaligned Huge Pages in Public Cloud Virtual Machines
    Mani Tofigh (Hofstra University & Columbia University), Edward Guo (Columbia University), Zhaoxi Shi, Weiwei Jia (The University of Rhode Island), Xiaoning Ding (New Jersey Institute of Technology), Neil Zhao (UT Austin & NVIDIA), Asaf Cidon (Columbia University), Jianchen Shan (Hofstra University)
  2. 16:40
    Access-Rate-Aware Hugepage Management at Warehouse Scale
    Mihailo Rancic (Carnegie Mellon University), Vaibhav Gogte, Nilay Vaish, Chris Kennelly, Patrick Xia (Google), Akshitha Sriraman (Carnegie Mellon University)
  3. 17:00
    FairMem: Enabling Performance Fairness for Colocated Memory-Intensive Applications on Tiered Main Memory Servers
    Amanda Raybuck (University of California, Davis), Wei Zhang (University of Connecticut), Ali Mansoorshahi (UT Austin), Aditya Kamath (University of Washington), Yojan Chitkara (UT Austin), Yuyi Li (University of California, Davis), Orhan Unuvar, Mattan Erez (UT Austin), Simon Peter (University of Washington)
  4. 17:20
    Interrupt-Free I/O Event Handling
    Yuki Tsujimoto (Keio University), Kenichi Yasukata (IIJ Research Laboratory), Kenta Ishiguro (Inria), Kenji Kono (Keio University)

Day 2 · Thursday, November 19, 2026

Time
Track A · Siloso 2
Track B · Siloso 3
09:00 – 10:30
Keynote 2: Joseph M. HellersteinSiloso 2 + 3
10:30 – 11:15
Coffee Break
11:15 – 12:35
Track A · Siloso 2
Session 2A: Datacenter Networking
Chair: Gyuyeong Kim
  1. 11:15
    FCP: Fast User-space RDMA Control Plane Setup for Elastic Computing
    Junxue Zhang, Han Tian (University of Science and Technology of China), Xinyang Huang, Wenxue Li, Kaiqiang Xu (Hong Kong University of Science and Technology), Dian Shen (Southeast University), Kai Chen (Hong Kong University of Science and Technology)
  2. 11:35
    RECAP: Authority-Aware Congestion Control for Packet-Sprayed RDMA Networks
    Leixin Cai, Jiao Zhang, Yuxiang Wang, Xiaolin Wang, Haoyu Zhang, Huimin Luo (Beijing University of Posts and Telecommunications)
  3. 11:55
    FluxReduce: Straggler-Resilient In-Network Aggregation for Distributed Machine Learning
    George Karlos, Henri E. Bal (Vrije Universiteit Amsterdam), Lin Wang (Paderborn University)
  4. 12:15
    SlimeMold: Scaling Distributed Hardware Load Balancers via a Unified Giant Connection Table
    Ziyuan Liu (Beihang University & Microsoft Research), Zhixiong Niu, Ran Shu (Microsoft Research), Jianqin Zhang, Guohong Lai, Liang Gao (Ruijie Networks), Na Wang, Zongying He (Broadcom Inc.), Jacob Nelson, Dan R. K. Ports (Microsoft Research), Lihua Yuan (Microsoft), Peng Cheng, Yongqiang Xiong (Microsoft Research)
Track B · Siloso 3
Session 2B: LLM Serving (Short Papers)
Chair: TBA
  1. 11:35
    On Evaluating Performance of LLM Inference Serving SystemsShort
    Amey Agrawal (Georgia Tech), Nitin Kedia (UT Austin), Anmol Agarwal (Apple Inc), Jayashree Mohan, Nipun Kwatra (Microsoft Research), Souvik Kundu (Intel Research), Ramachandran Ramjee (Microsoft Research), Alexey Tumanov (Georgia Tech)
  2. 11:50
    SLO-Aware Proactive History Compression for Long-History Generative Recommendation ServingShort
    Xiaoyu Ma, Weihao Cui (Shanghai Jiao Tong University)
  3. 12:05
    Revisiting Cauchy-Schwarz Norm Bounds for SDC Detection in LLM InferenceShort
    Anton Altenbernd, Dylan Everingham, Philipp Wiesner, Odej Kao (TU Berlin)
  4. 12:20
    Profile-Guided Intent-Based Selection for Edge LLM ServingShort
    Arda Goknil, Hui Song, Erik Johannes Husom, Rustem Dautov, Merve Astekin, Dumitru Roman (SINTEF), Valerio Frascolla (Intel)
12:35 – 14:00
Lunch
14:00 – 15:20
Track A · Siloso 2
Session 2C: LLM Serving: Scheduling & Autoscaling
Chair: Haiying Shen
  1. 14:00
    No Request Left Behind: Tackling Heterogeneity in Long-Context LLM Inference with Medha
    Amey Agrawal (Georgia Tech), Haoran Qiu (Microsoft Azure Research), Junda Chen (University of California San Diego), Íñigo Goiri (Microsoft), Rayyan Shahid (Georgia Institute of Technology), Ram Ramjee (Microsoft Research India), Alexey Tumanov (Georgia Tech), Esha Choukse (Microsoft)
  2. 14:20
    CUCUMIS: Request-Sensitive First-Token Scheduling for Shared GPU LLM Serving
    Diaohan Luo (University of Chinese Academy of Sciences), Heng Wu, Yuewen Wu, Zhen Tang (Institute of Software, Chinese Academy of Sciences), Zichao Yang (University of Chinese Academy of Sciences), Ligang Ren (Beijing Branch of China United Network Communication Co., Ltd), Wenbo Zhang (Institute of Software, Chinese Academy of Sciences)
  3. 14:40
    TokenPipe: Stage-aware Autoscaling for Disaggregated LLM Serving with Token Velocity
    Ruiqi Lai, Chengzhi Lu, Hongrui Liu (NTU Singapore), Siyu Cao (Singapore Management University), Li Leqi (Huawei), Zonghao Liu (NTU Singapore), Siyang Shao (Georgia Institute of Technology), Yixin Zhang (Century Frontier Investment), Luo Mai (University of Edinburgh), Dmitrii Ustiugov (NTU Singapore)
  4. 15:00
    Ceres: Low-latency Multi-Agent LLM Serving via Perceiving Execution Disparity across Agents
    Jinyuan Chen (Shanghai Jiao Tong University), Jiuchen Shi (Shanghai Jiao Tong University & The Hong Kong Polytechnic University), Quan Chen, Weihao Cui (Shanghai Jiao Tong University), Jie Wu (Cloud Computing Research Institute, China Telecom & Temple University), Minyi Guo (Guizhou Provincial Laboratory of Big Data, College of Computer Science and Technology, Guizhou University & Shanghai Jiao Tong University)
Track B · Siloso 3
Session 2D: Storage & Data Management
Chair: TBA
  1. 14:00
    CoFDP: User-Kernel Cooperative Storage for Flexible Data Placement SSDs
    Sungjin Byeon, Jaewan Park, Soon Hwang, Sangyun Lee, Kyungwook Min, Hajin Moon, Taewan Noh (Sogang University, Seoul), Joo-Young Hwang, Junyoung Han (Samsung Electronics), Youngjae Kim (Sogang University, Seoul)
  2. 14:20
    Making Sense of DPU Performance for Cloud Data Processing
    Jiasheng Hu, Chihan Cui, Yuanfan Chen (University of Toronto), Philip A. Bernstein (Microsoft Research), Jialin Li (NUS), Qizhen Zhang (University of Toronto)
  3. 14:40
    Spanner-RSS+: Diagnosing and Reducing Safe-Time Waiting of Read-only Transactions in Spanner-RSS
    Sejong Kim, Hena Cho, Yon Dohn Chung (Korea University)
  4. 15:00
    Fine-Grained Data Provenance Through User-Defined Functions in Spark
    Muhammad Ali Gulzar (Virginia Tech)
15:20 – 16:05
Coffee Break
16:05 – 17:25
Track A · Siloso 2
Session 2E: Observability & Reliability
Chair: TBA
  1. 16:05
    Two Parents Are Better Than One: Enhancing Critical Path Analysis for Distributed Tracing with CPA2
    Julia Cusatis (The Pennsylvania State University), Jingyuan Chen (Princeton University), Timothy Zhu (The Pennsylvania State University)
  2. 16:25
    Crinkle: Tenant-Driven Debugging on Cloud Testbeds
    Alexander Wolosewicz (Illinois Tech), Raffay Atiq (SRI), Nishanth Shyamkumar, Van Anderson (Illinois Tech), Vinod Yegneswaran, Ashish Gehani (SRI), Nik Sultana (Illinois Tech)
  3. 16:45
    Seeing is believing: The case for hardware-centric inefficiency detection and remediation
    Ziji Chen (University of Oxford), Steven W. D. Chien (University of St Andrews), Peng Qian, Noa Zilberman (University of Oxford)
  4. 17:05
    Panacea: From Ambiguous Incidents to Safe Mitigation Policies at Azure ScaleIndustry
    Ze Li (Microsoft Azure), Chang Lou (University of Virginia), Vivek Ramamurthy, Vignatha Yenugutala, Eion Blanchard, Minghua Ma, Hemant Malik, Shweta Patil, Murali Chintalapati (Microsoft Azure)
Track B · Siloso 3
Session 2F: Quantum Computing & Emerging Hardware
Chair: TBA
  1. 16:05
    TransCache: Opportunistic Pre-Transpilation Caching for Low-Latency Quantum-as-a-Service
    Xinyi Li, Yifeng Peng (Stevens Institute of Technology), Siyuan Niu (University of Central Florida), Zhiding Liang (Rensselaer Polytechnic Institute), Hao Wang, Ying Wang (Stevens Institute of Technology)
  2. 16:25
    QuShare: A Multi-Tenant QPU Sharing Method for Hybrid Quantum-Classical Workflows on Quantum Cloud Platforms
    Jiwon Ha (Seoul National University), Sookyoung Park (Sookmyung Women’s University), Hyeonsang Eom (Seoul National University), Yoonhee Kim (Sookmyung Women’s University)
  3. 16:45
    Characterizing and Mitigating Temporal Noise Drift in Quantum Reservoir Computing
    Aditya Ranjan (Northeastern University), Tirthak Patel (Rice University), Devesh Tiwari (Northeastern University)
  4. 17:05
    FAHO: A Time-Sharing System for PIM-based Applications
    Maxime Collette, Weihao Ni (Inria), Eric Okala, Renaud Lachaize, Alain Tchana (Université Grenoble Alpes)

Day 3 · Friday, November 20, 2026

Time
Track A · Siloso 2
Track B · Siloso 3
09:00 – 10:20
Track A · Siloso 2
Session 3A: ML for Systems & Networks
Chair: TBA
  1. 09:00
    (Near-)Pareto-Optimal Synthesized OS-Level Scheduling for Serverless Systems
    Yuxuan Zhao (Leiden University), Zhao Yang (VU Amsterdam), Weikang Weng (LIACS, Leiden University), Vincent François-Lavet (VU Amsterdam), Rob van Nieuwpoort (Leiden University), Alexandru Uta (DFINITY)
  2. 09:20
    SuperCA: Hierarchical Reinforcement Learning for Scalable Compiler Phase Ordering on HPCIndustry
    Guohui Zhang (Northwest University), Xiaoyang Sun (ACE3 AI LTD), Zhanhong Ding, Sai Liu (Northwest University), Pengyu Huang (Xidian University), Wei Wang (Northwest University), Zheng Wang (University of Leeds)
  3. 09:40
    Incisor: Ex Ante Instance Selection in the Cloud
    Michael Laurenzano, Shihan Cheng, David Hyde (Vanderbilt University)
  4. 10:00
    FlowWise: Fast–Slow Model Orchestration for Line-Rate Network Traffic Intelligence
    Shinan Liu (University of Hong Kong), Ted Shaowang (University of Chicago), Gerry Wan (Stanford University), Jeewon Chae, Jonatas Marques, Sanjay Krishnan, Nick Feamster (University of Chicago)
Track B · Siloso 3
Session 3B: Visions for the Cloud (Short/Vision Papers)
Chair: Anjali
  1. 09:20
    Scale-Up-First: The Cloud as One Logical ComputerVision
    Daniel Barcelona-Pons, Pedro Garcia Lopez, Stepan Klymonchuk (Universitat Rovira i Virgili)
  2. 09:35
    The End of Time (as a Cloud Billing Model)Vision
    Tom Kuchler, Saurav Banka, Ana Klimovic (ETH Zurich)
  3. 09:50
    The Sorry State of Stateful ScalingVision
    Chirag C Shetty (University of Illinois Urbana-Champaign, USA), Larisa Shwartz, Chandra Narayanaswami, Saurabh Jha (IBM Research), Indranil Gupta (University of Illinois Urbana-Champaign, USA)
  4. 10:05
    Load Fiction: How Your Load Generator is Misleading YouShort
    Amit Das, Anant Goyal, Yug Mittal (University of Illinois Urbana-Champaign, USA), Saurabh Jha (IBM Research), Chirag C Shetty, Indranil Gupta (University of Illinois Urbana-Champaign, USA)
10:20 – 11:00
Coffee Break
11:00 – 12:20
Track A · Siloso 2
Session 3C: LLM Inference: KV Cache & Attention
Chair: TBA
  1. 11:00
    TENURE: Residency Leases for Tiered LLM KV Caches
    Amir Noohi, Antonio Barbalace (University of Edinburgh)
  2. 11:20
    MosaicKV: SLO-Aware KV-Cache Virtualization for Multi-Tenant LLM Serving in the Cloud
    Rui Li, Shuang Cao (Hill Research)
  3. 11:40
    FastDC: Fast KV Dimensionality Compression for Decompression-Free and Attention-Accelerated LLM Inference
    Zeyu Zhang, Haiying Shen (University of Virginia)
  4. 12:00
    EDiff: A Diffusion LLM Inference Engine with Efficient Computation and KV Cache Management
    Ye Tian, Juntao Zhao (The University of Hong Kong), Zhenkun Cai (AWS AI), Zhen Jia (Amazon Web Services, Inc.), Yida Wang (Amazon), Chuan Wu (The University of Hong Kong)
Track B · Siloso 3
Session 3D: Latency-Critical Services & Microservices
Chair: Timothy Zhu
  1. 11:00
    Spandana: Reconciling Strict SLOs with Low Cost under Fine-Grained Load Fluctuations
    Dilina Dehigama, Shyam Jesalpura, Zeyu Xu, Marton Nemeth, Shengda Zhu (University of Edinburgh), Marios Kogias (Imperial College London), Boris Grot (University of Edinburgh)
  2. 11:20
    Tamo: Taming Overload of Highly Dynamic Microservices via Precise Call Graph Prediction and Workload-Aware Scheduling
    Yaoxuan Li, Pu Pang, Zhengxuan Yan, Jiuchen Shi, Quan Chen, Minyi Guo (Shanghai Jiao Tong University)
  3. 11:40
    Helix: A Flexible and Scalable NextGen Cellular Core
    Bilal Saleem, Omar Basit (Purdue University), Jiayi Meng (University of Texas at Arlington), Ajay Lotan Thakur, Iftekhar Alam, Christian Maciocco (Intel), Y. Charlie Hu (Purdue University), Larry Peterson (Princeton University), Muhammad Shahbaz (University of Michigan)
  4. 12:00
    SkyAccel: A Cross-Cloud Acceleration System for Latency-Sensitive WorkloadsIndustry
    Matthew Yang Liu (Bilibili), Hao Che (The University of Texas at Arlington), Schahram Dustdar (ICREA)
12:20 – 13:50
Lunch
13:50 – 15:20
Track A · Siloso 2
Session 3E: Agentic AI Systems
Chair: Dmitrii Ustiugov
  1. 13:50
    Cadence: Taming Coupled Contention in Disaggregated Agent Clouds
    Rui Li, Shuang Cao (Hill Research)
  2. 14:10
    Laplace: Predictive Sandbox Orchestration for LLM Agents
    Yihui Zhang, Tianyu Wo, Jinghao Wang (Beihang University), Xiaoyang Sun (University of Leeds), Menghao Zhang, Cangzhou Yuan, Li Li, Chunming Hu (Beihang University), Albert Y. Zomaya (The University of Sydney), Renyu Yang (Beihang University)
  3. 14:30
    When, Which, and What to Fork: The Launch-Policy Design Space for LLM AgentsShort
    Zirui Wang, Mohamed M. Kassem, Adam Barker (Huawei Edinburgh Research)
  4. 14:45
    Rethinking AI Cloud Infrastructure for Agentic Serving Systems with the Artemis Experimentation FrameworkVision
    JooYoung Park, Leonid Kondrashov, Boxi Zhou, Hongrui Liu, Zonghao Liu, Chengzhi Lu (NTU Singapore), Riccardo Mancini (AWS), Esha Choukse (Microsoft), Haris Javaid, German Sviridov (AMD), Tao Peng, Chen Zhao (Ant Group), Marios Kogias (Imperial College London), Luo Mai (University of Edinburgh), Anastasia Avdeeva, Aleksei Gusev (NCSpeech), Dmitrii Ustiugov (NTU Singapore)
  5. 15:00
    Evaluating Agentic Systems Across the Lifecycle: A Taxonomy of Drifts and Success CriteriaIndustry
    Gaurav Saxena (Independent Researcher), Ildar Iskhakov, Innokentii Konstantinov, Matvey Kukuy (Archestra.AI)
Track B · Siloso 3
Session 3F: Energy & Sustainability
Chair: Walid Abdelrahman Hanafy
  1. 13:50
    Flex: Grid-Responsive Provisioning and Scheduling for Elastic Cloud Clusters
    Walid Abdelrahman Hanafy, Li Wu, David Irwin, Prashant Shenoy (University of Massachusetts Amherst)
  2. 14:10
    XWind: A Cross-site Router for Large Language Model Inference Serving at Renewable Energy Farms
    Tella Rajashekhar Reddy, Atharva Deshmukh, Liangcheng Yu, Chaojie Zhang, Mike Shepperd, Rohan Gandhi, Anjaly Parayil, Srinivasan Iyengar, Ajay Manchepalli, Debopam Bhattacherjee (Microsoft)
  3. 14:30
    Reconciling Bottom-Up Metrics with Top-Down Reporting for Carbon Accounting in the CloudVision
    Philipp Wiesner (TU Berlin, Logsight.ai), Loïc Lannelongue (University of Cambridge), Alexander Acker (Logsight.ai), Odej Kao (TU Berlin)
  4. 14:45
    WattOS: Toward a Power Control Plane for AI Data CentersVision
    Pu Zhang, Guoming Tang (The Hong Kong University of Science and Technology (Guangzhou))
  5. 15:00
    PEACE: Power and Performance Aware Colocation for Efficient GPU Spatial PartitioningShort
    Bing-Shiun Han, Chaitanya Sundeep Subhedar, Kunaal Parekh, Wan-Chu Lin, Zhenhua Liu, Anshul Gandhi (Stony Brook University)
15:20 – 15:50
Coffee Break
15:50 – 16:50
Track A · Siloso 2
Session 3G: Federated & Edge Learning
Chair: TBA
  1. 15:50
    FedRD: Towards Memory-efficient Federated Learning via Adaptive Recomputation and Defragmentation
    Kahou Tam, Ziyu Gong, Li Li, ChengZhong Xu (University of Macau)
  2. 16:10
    Memory-Decoupled Layer-Wise Fine-Tuning for Efficient On-Device LLM Adaptation
    Yebo Wu, Jingguang Li, Chunlin Tian, Kahou Tam (University of Macau), Zhijiang Guo (The Hong Kong University of Science and Technology (Guangzhou)), Li Li (University of Macau)
  3. 16:30
    DynSL: Resource-Adaptive Model Training across Heterogeneous Edge Devices
    Boris Radovič (KAUST, University of Ljubljana), Istemi Ekin Akkus (Nokia Bell Labs), Marco Canini (KAUST), Ruichuan Chen, Alice Dethise, Ivica Rimac (Nokia Bell Labs), Veljko Pejović (University of Ljubljana)
Track B · Siloso 3
Session 3H: Vector Search & RAG
Chair: TBA
  1. 15:50
    NemoANN: I/O-Efficient Filtered Vector Search on SSDs with Graph Tunneling
    Nakyung Lee, Soobin Cho, Jiwoong Park (Sungshin Women's University), Gyuyeong Kim (Yonsei University)
  2. 16:10
    Ghost Vectors: Storage-Layer Reconstruction and Cryptographic Erasure in HNSW Vector Databases
    Chandranil (Nil) Chakraborttii, Jackeline García Alvarado, Sitora Abdulofizova, Talakinet Wabelo, Shivanshu Dwivedi (Trinity College, USA)
  3. 16:30
    OpRAG: A Resource-Deterministic Runtime for GPU-Backed Multi-Stage RAG Workflows
    Arup Kumar Sarker, Mills Staylor (University of Virginia), Aymen Alsaadi (Rutgers University), Gregor von Laszewski (Biocomplexity Institute and Initiative), Yue Cheng (University of Virginia), Shantenu Jha (Rutgers University), Geoffrey Fox (University of Virginia)
16:50 – 17:20
Closing Remarks
All times are local (Singapore, UTC+8). The program is subject to change.