We are the Distributed Data Processing Systems Laboratory (DDPS Lab) at Hanyang University. We build high-performance, cost-effective, and easy-to-use computer systems, from edge devices to cloud-scale clusters.
| Area | Focus |
|---|---|
| ☁️ Cloud Computing | Spot instance datasets and resource optimization, spot instances for LLM training and serving, cloud performance modeling and cost optimization |
| 🧠 System for AI | LLM serving on heterogeneous GPUs, WebAssembly runtimes for AI workloads, edge-cloud continuum, AI agent infrastructure |
| 🛠️ AI for System | Root cause analysis for cloud systems, AI agents for system operations, AIOps |
| 📊 BigData Processing | Time-series processing, serverless data pipelines, large-scale matrix operations |
| Project | Description | Paper |
|---|---|---|
| SpotLake | Public spot instance dataset archive across AWS, Azure, and GCP | IISWC '22, WWW '23 Demo |
| KubePACS | Kubernetes cluster on performant, highly available, cost-efficient spot instances | Middleware '26 |
| ShuntServe | Cost-efficient LLM serving on heterogeneous spot GPU clusters | FGCS |
| SpotVista | Availability-aware multi-node spot instance recommendation | |
| LiveMigrate-Detector | Workload-aware live-migratable cloud instance detector | CCGrid '24, TCC |
| PROFET | Profiling-based CNN training latency prediction for GPU cloud instances | IEEE BigData '22 |
| FunctionBench | Workload suite for serverless cloud function services | IEEE CLOUD '19 |
| DoS | Dense or Sparse: elastic SpMM for big-data processing | IEEE TBD |
See all projects at ddps.cloud/project.
📮 ddpslab@hanyang.ac.kr
🏫 Department of Data Science · Department of Artificial Intelligence, Hanyang University, Seoul
Prospective students and collaborators are welcome to reach out by email.
한양대학교 분산데이터처리시스템연구실· DDPS Lab · Hanyang University · ddps.cloud