All program times are in Central European Summer Time (CEST, UTC+2).
Tutorial 3: Energy Efficiency Benchmarking of GPU Servers - Why It’s Challenging, What Matters, and How to Do It Right### Wednesday, May 6, 2026
| Time |
Slot |
Program |
| 08:00-08:45 |
|
Registration |
| 08:45-09:00 |
|
Conference Opening |
| 09:00-10:00 |
Keynote 1 |
Leana Golubchik (University of Southern California) — Systems for AI: Predicting Performance of Machine Learning Workloads |
| 10:00-10:30 |
Session 1 |
Energy Efficiency & Simulation Performance
- 10:00-10:20 Improving Energy Efficiency and Performance of Weather and Climate Simulations by Leveraging the Heterogeneity of Modern Systems
Julius Plehn, Christian von Elm, Pay Giesselmann, Carsten Clauss, Hendryk Bockelmann, Robert Schöne, Jan Frederik Engels(Research Track)
- 10:20-10:30 Cross-Platform, Cross-Framework Development of Hybrid-Parallel Matrix-Multiplication codes
Vyuhita Bonthu, Nikhil Hegde (Short – Research Track)
|
| 10:30-11:00 |
|
Coffee Break |
| 11:00-12:36 |
Session 2 |
Cloud Systems & Resource Efficiency
- 11:00-11:20 CarbonShare: Carbon-Fair Allocation for Shared Clusters
John Thiede, David Irwin, Prashant Shenoy(Research Track)
- 11:20-11:40 Kill Smart, Run Fast: Using Job Termination for Efficient and Fair Scheduling in Data Centers
Rostislav Razumchik, Andrea Marin, Adityo Anggraito (Research Track)
- 11:40-12:00 Understanding Foundational Library Energy Consumption
Jacob D. Hauenstein, Timothy S. Newman (Research Track)
- 12:00-12:12 The Impact of Memory Configuration on Server Efficiency
Maximilian Meissner, Khang Pham, Aaron Cragin, Klaus-Dieter Lange, Samuel Kounev(Industry Track - Best Industry Paper candidate)
- 12:12-12:24 Evaluating Kubernetes Performance for GenAI Inference: From Automatic Speech Recognition to LLM Summarization
Sai Sindhur Malleni, Raúl Sevilla, Aleksei Vasilevskii, José Castillo Lema, André Bauer(Industry Track - Best Industry Paper candidate)
- 12:24-12:36 A Transparent and Efficient Performance Analysis Approach to Enhance DPDK Observability
Adel Belkhiri, Arnaud Fiorini, Matthew Khouzam, Heng Li (Industry Track - Best Industry Paper candidate)
|
| 12:36-14:00 |
|
Lunch / Posters & Demonstrations |
| 14:00-15:32 |
Session 3 |
AI & LLM Performance
- 14:00-14:20 SweetSpot: An Analytical Model for Predicting Energy Efficiency of LLM Inference
Hiari Pizzini Cavagna, Andrea Proia, Giacomo Madella, Giovanni Battista Esposito, Francesco Antici, Daniele Cesarini, Zeynep Kiziltan, Andrea Bartolini (Research Track)
- 14:20-14:40 B-Perf: Black-box Performance Antipattern Detection Using System-level Execution Tracing
Morteza Noferesti, Mahsa Panahanddeh, Naser Ezzati-Jivan (Research Track)
- 14:40-15:00 ORION: Integrated Runtime Modelling for Predicting Deep Learning Training Time
Alireza Pourali, Hamzeh Khazaei (Research Track)
- 15:00-15:20 SwiftSNNI: Optimized Scheduling for Secure Neural Network Inference (SNNI) on Multi-Core Systems
Kanwal Batool, Saleem Anwar, Francesco Regazzoni, Andy Pimentel, Zoltán Ádám Mann(Research Track)
- 15:20-15:32 On the Efficiency and Disruption Trade-Offs of Kubernetes Packing Heuristics
Mariane Santos Zeitouni, Oscar Brito, Matheus Rocha, Thiago Emmanuel Pereira, Gabriel Gomes (Industry Track)
|
| 15:32-16:00 |
|
Coffee Break |
| 16:00-17:00 |
Panel |
Performance engineering in the era of GenAI: what stays, what goes, and what's next (confirmed panelists: Leana Golubchik, Bjarne Stroustrup) |
| 17:00-17:30 |
Session 4 |
System Observability & Latency
- 17:00-17:20 Benchmarking the Overhead of Distributed Tracing Agents
David Georg Reichelt, Shinhyung Yang, Marcel Hanson, Wilhelm Hasselbring(Research Track)
- 17:20-17:30 Modeling Extreme End-to-End Delays for Availability Assessment on Latency Datasets
Orangel Azuaje Contreras, Ana Aguiar(Short – Research Track)
|
| Evening |
|
Welcome Reception |
| Time |
Slot |
Program |
| 08:30-09:00 |
|
Registration |
| 09:00-10:00 |
Keynote 2 |
Didem Unat (Koç University) — Illuminating Multi-GPU Communication Paths |
| 10:00-10:30 |
Session 5 |
Benchmarking & Profiling Infrastructure
- 10:00-10:20 benchkit: A Declarative Framework for Composable Performance Evaluation of System Software
Antonio Paolillo, Mats Van Molle, Ken Hasselmann(Research Track)
- 10:20-10:30 Performance and Cost Implications of Migrating Serverless Functions from x86 to ARM based Servers
Yassine Lazreg, Saman Akbari, Manfred Hauswirth (Short – Research Track)
|
| 10:30-11:00 |
|
Coffee Break |
| 11:00-12:30 |
Session 6 |
Performance Modeling and Optimization of Complex Systems
- 11:00-11:20 Variability-Guided Performance Optimization
Eitan Frachtenberg, Viyom Mittal, Mohammed Baydoun, Aditya Dhakal, Izzat El Hajj, Dejan Milojicic(Research Track - Best Paper candidate)
- 11:20-11:40 Energy-Efficient Right-Sizing of Kafka-like Message Brokers for IoT Workloads
Govind KP, Romain Rouvoy, Guillaume Pierre (Research Track - Best Paper candidate)
- 11:40-12:00 Performance Analysis and Optimization of 3D Generative Diffusion Models across GPU
Jeeho Ryoo, Yongchan Jung, Muhammad Ali Khaliq, Weidong Zhang, Jiatong Han, Byeong Kil Lee(Research Track - Best Paper candidate)
- 12:00-12:20 A Comparative Evaluation of Imputation Models for Agricultural Weather Networks
Awanish Khanal, Monowar Hasan(Research Track - Best Paper candidate)
|
| 12:30-14:00 |
|
Lunch / Posters & Demonstrations |
| 14:00-15:28 |
Session 7 |
GPU & Heterogeneous Computing
- 14:00-14:20 MQGPU: A Multi-Queue Scheduling Framework For GPU Accelerated Serverless Functions
Alexander Fuerst, Siddharth Anil, Prateek Sharma (Research Track)
- 14:20-14:40 LSTC: Large-Scale Triangle Counting on Single GPU
Kishan Tamboli, Vishwesh Jatala (Research Track)
- 14:40-15:00 Pulse: A Profiling and Visualization Infrastructure for Heterogeneous Managed Systems
Michail Papadimitriou, Maria Xekalaki, Orion Papadakis, Ruiqi Ye, Athanasios Stratikopoulos, Christos Kotselidis (Research Track)
- 15:00-15:12 Low-Latency ML Offloading Across Edge and IoT Devices
Konstantinos Papazafeiropoulos, Anastasia Mallikopoulou, Anastassios Nanos, Georgios Goumas, Nectarios Koziris(Industry Track)
- 15:12-15:20 Energy- and Quantization-aware DNN Partitioning in the Edge-Cloud Continuum
S. Nicosanti, G. Russo, V. Cardellini (Emerging Research Track)
- 15:20-15:28 A Taxonomy of Application Properties for Mixed-Precision Autotuning
G. Gedik, C. von Elm, R. Schöne (Emerging Research Track)
|
| 15:28-16:00 |
|
Coffee Break |
| 16:00-17:30 |
Session 8 |
Adaptive Cloud & Edge
- 16:00-16:20 WASL: Harmonizing Uncoordinated Adaptive Modules in Multi-Tenant Cloud Systems
Ahsan Pervaiz, Anwesha Das, Vedant Kodagi, Muhammad Husni Santriaji, Henry Hoffmann (Research Track)
- 16:20-16:32 KLUE: A Framework for Cost-Effective Experimentation in Emulated Kubernetes Clusters
Kayky Fidelis, Geraldo Junior, Caetano Albuquerque, Giovanni Farias, Thiago Emmanuel Pereira, Fabio Morais, Kilian Melcher (Industry Track)
- 16:32-16:42 FLYT: Transparent and Elastic GPU Provisioning for Multi-Tenant Cloud Services
Santhosh Kumar M, Sameer Ahmad, Armaan Chowfin, Purushottam Kulkarni, Anand Eswaran, Praveen Jayachandran(Short – Research Track)
- 16:42-16:52 To Offload or Not To Offload: Model-driven Comparison of Edge-native and On-device Processing in the Era of Accelerators
Nathan Ng, David Irwin, Ananthram Swami, Don Towsley, Prashant Shenoy (Short – Research Track)
- 16:52-17:30 SPEC Research Group presentation; SPEC Kaivalya Dixit Distinguished Dissertation Award talk; ICPE MIP Award presentation
|
| Evening |
|
Conference Dinner |
| Time |
Slot |
Program |
| 08:30-09:00 |
|
Registration |
| 09:00-10:00 |
Keynote 3 |
Jeff Hammond (NVIDIA) — State-of-the-Art Communication Software for Supercomputers and Its Applications |
| 10:00-10:32 |
Session 9 |
Java & Heap Performance
- 10:00-10:20 MapReplay: Trace-Driven Benchmark Generation for Java HashMap
Filippo Schiavio, Andrea Rosa, Júnior Löff, Lubomír Bulej, Petr Tuma, Walter Binder (Research Track)
- 10:20-10:32 G1HeapVis: Visualizing and Measuring Heap Fragmentation
Oleksandr Kachur (Industry Track)
|
| 10:32-11:00 |
|
Coffee Break |
| 11:00-12:30 |
Session 10 |
Adaptive Systems & Predictive Management
- 11:00-11:20 Are We There Yet? Predicting if Executing Applications are Near Completion
Mohammad Sonji, Mohammed Baydoun, Safaa Diab, Amir Nassereldine, Pedro Bruel, Aditya Dhakal, Rolando Pablo Hong Enriquez, Gourav Rattihalli, Diman Zad Tootaghaj, Gallig Renaud, Barbara Chapman, Fatima K. Abu Salem, Eitan Frachtenberg, Dejan Milojicic, Izzat El Hajj(Research Track)
- 11:20-11:40 Holpaca: Holistic and Adaptable Cache Management for Shared Environments
José Pedro Peixoto, Alexis Gonzalez, Janki Bhimani, Raju Rangaswami, Cláudia Brito, João Paulo, Ricardo Macedo (Research Track)
- 11:40-12:00 Energy-efficient Dynamic Partitioning and Tensors Compression of AI Applications in Smart Eyewears
Abednego Wamuhindo Kambale, Samin Shokrivahed, Giacomo Verticale, Francesca Palermo, Diana Trojaniello, Danilo Ardagna (Research Track)
- 12:00-12:20 An Evaluation Study of Generative AI Systems: Framework-Aware Performance Under Real-World Constraints
Abed Matinpour, Farhoud Jafari Kaleibar, Sara Fehresti, Shaylin Ziaei, Marin Litoiu (Research Track)
- 12:20-12:30 Trust Your Local Scaler: A Continuous, Decentralized Approach to Autoscaling
Martin Straesser, Stefan Geissler, Stanislav Lange, Lukas Kilian Schumann, Tobias Hossfeld , Samuel Kounev(Journal First Track)
|
| 12:30-14:00 |
|
Lunch / Posters & Demonstrations |
| 14:00-15:30 |
Session 11 |
Emerging Trends & Data Challenges
- 14:00-14:08 Leveraging LLMs for Structured Information Extraction and Analysis from Cloud Incident Reports
X. Chu, S. Ilager, Y. Zang, S. Talluri, A. Iosup (Emerging Research Track)
- 14:08-14:16 Detecting Silent Failures in Multi-Agentic AI Trajectories
D. Pathak, F. George, H. Kumar, A. Roy, K. Ray, M. Verma, P. Moogi(Emerging Research Track)
- 14:16-14:24 Unsupervised Cycle Detection in Agentic Applications
F. George, D. Pathak, H. Kumar, K. Ray, M. Verma, P. Moogi(Emerging Research Track)
- 14:24-14:32 An Agent-Based Approach to Automating Software Performance Testing
E. Binder, X. Li, A. Janes(Emerging Research Track)
- 14:32-14:40 Representation-Aware RCA with Large Language Models
Y. Wen, M. Panahanddeh, M. Chouchen, W. Hamou-Lhadj(Emerging Research Track)
- 14:40-14:48 Platooning Without Leaders: A Performance-Driven Reframing of Cooperative Vehicle Systems
Jai Aakaash J.S, A. Arulappan (Emerging Research Track)
- 14:48-14:56 A Bayesian Way of Estimating Method Cost from Conflicting Profiler Data
N. Couderc (Emerging Research Track)
- 14:56-15:04 Beyond Reproduction: Uncovering Latent Performance Regressions with LLM-Guided Fuzzing
R. Zheng, Y. Zhao, L. Xiao, W. Shang, L. Liao (Emerging Research Track)
- 15:04-15:12 Performance Alert Triage with Time-Aware Learning and Multi-Scale Time-Series Features
Adem Hmissa, Rodolphe Laurent Louis Picot, Rani Naaman, Ahmad Shahnejat Bushehri, Rim Zrelli, Felipe Gohring de Magalhaes, Gabriela Nicolescu(Data Challenge Track)
- 15:12-15:20 Performance Regressions Prediction using Time Series Classification: A Case Study
Federico Di Menna, Luca Traini(Data Challenge Track)
- 15:20-15:28 Weekly Seasonality in Cloud Demand: Lessons from Snowflake’s Shaved Ice Dataset
Hector Pena, Steven Cheun(Data Challenge Track)
|
| 15:30-16:00 |
Closing |
Closing Session |