본문 바로가기 메뉴 바로가기

SNU Professors Jae Wook Lee and Byung-Gon Chun Present Their Research Papers at 2019 USENIX Annual Technical Conference

  • Uploaded by

    관리자

  • Upload Date

    2019.06.24

  • Views

    339

SNU Professors Jae Wook Lee and Byung-Gon Chun Present Their Research Papers at 2019 USENIX Annual Technical Conference


▲ Professor Jae Wook Lee, Professor Byung-Gon Chun

 
SNU College of Engineering (Dean Kookheon Char) announced that Professor Jae Wook Lee and Professor Byung-Gon Chun of the Department of Computer Science and Engineering have submitted three research articles for the 2019 USENIX Annual Technical Conference.
 
Firstly, Professor Lee’s team presented two researches that can significantly reduce the latency of SSD-based big data applications that are run at the data center. The team developed a low-cost, highly-reliable NAND block write/erase method that greatly shortened the tail latency of NAND flash memory based SSD. With this new technology, the maximum read-and-response time drops from 10ms to only few hundred μs.
 
They also cooperated with the research team led by Professor Jinkyu Jeong of the Sungkyunkwan University to complete an asynchronous OS kernel I/O stack that can reduce the I/O latency of the ultra-low latency SSD. This research significantly reduced the read latency of storage device; thus, an anticipation that the satisfaction rate of data center users will greatly rise.
 
On the other hand, Professor Chun’s team suggested the Apache Nemo, an optimized system that supports various environment and data properties using the distributed processing of big data application. Professor Chun is currently operating the Apache Nemo project, which was approved by the Apache open source incubator of the Apache Software Foundation (ASF). 
 
Apache Nemo is recognized as a next-generation big data analysis technology for providing an interface to easily create policies to optimize distributed processing. The function of a complex system is improved when it is optimized to a specific environment and data property. Apache Nemo demonstrates higher performance than Apache Spark, which is most widely-used.
 
Meanwhile, Lee and Chun’s works are published on https://www.usenix.org/conference/atc19/technical-sessions and the final publications will be available starting from July. The 2019 USENIX Annual Technical Conference will be held in Renton, Washington this coming July.