|
|
| : |
|
| Special Talk : |
좌장 : |
|
 |
| 발표제목 : MangoBoost: Fast and Smart Networking Solutions for Next-Generation AI Infrastructure |
|
| 발표자 : 김장우 |
이메일 : |
|
| 소속 : 서울대학교/망고부스트 |
부서 : 전기정보공학부/경영진 |
|
| 직위 : 교수/CEO |
발표일시 : 2026. 1. 28.(수) 13:30-15:30 |
|
|
|
|
| 발표자약력 : |
|
(경력)
2022-현재 망고부스트 CEO
2017-현재 서울대학교 전기정보공학부 교수
2010-2016 포항공과대학교 컴퓨터공학과 교수
2009-2010 Oracle Corporation, Senior Engineer
2008-2009 Sun Microsystms, Technical Staff
(학력)
- Carnegie Mellon University, Electrical and Computer Engineering (PhD), 2008
- Cornell University, Computer Science (석사), 2001
- Cornell University, Computer Engineering (학사), 1997 |
|
|
| 강연요약 : |
|
| To realize cost-effective AI processing, modern high-performance computer systems are deploying an increasing number of high-performance devices (e.g., GPU, SSD, NPU, NIC) per server, and they all aim to maximize the target application’s overall performance while minimizing the system’s total cost of operation. For the purpose, the next-generation datacenters are deploying newly-designed fast and smart networking solutions to orchestrate various computer devices in the most cost-effective way. In this talk, I will first introduce the key challenges in developing the next-generation networking systems for cost-effective AI processing, and recent advances in that direction. Next, I will present how the state-of-the-art solutions from MangoBoost can achieve such design goals successfully. |
|
|
|
|
|
|