DDPS Lab.

Research Area

We develop high-performance, cost-effective, and easy-to-use computer systems

Cloud Computing

Cloud computing is a key technology for building scalable and cost-effective computing systems. Our research covers to develop systems that performs the latest workloads such as BigData Processing and Deep Learning quickly and cost-effectively in a cloud environment

Research on Spot Instances Dataset
Research on Spot Instances for DNN Training
Research on Performance Model for Cloud CNN Training

System for AI

In order to effectively develop ML/DL algorithms, a fast and scalable system must be supported. Our study focuses on developing and optimizing systems for cost-effective and high-performance of ML/DL training and inference workloads. To achieve the goal, we analyze various accelerator hardware and develop scheduling algorithms for distributed systems.

Research on Automated DNN Training Optimization
Research on SLO and Cost Aware Inference System
Research on Edge Inference System

BigData Processing

It is important to configure a system to efficiently process large volumes of data generated at high velocity. Our research focuses on identifying a combination of hardware and platforms optimized according to the characteristics of the BigData workloads.

Research on Large-scale MM Optimization
Research on Serverless BigData Processing

System Performance Analysis

System performance analysis is the first study to be conducted before developing a performance model or optimized system. Our study analyzes the latency, cost, and utilization of ML/DL, BigData, and cloud systems with various workloads.