Fast and Efficient Data Processing Methods for All-flash Distributed Storage System REVIEW

The era of bigdata and the evolution of fast storage devices cause a distributed storage system to face unpredictable challenges. In this talk, I am talking about an efficient storage management scheme and performance improvement method to overcome the challenges as to storage capacity and performance. The first topic I am going to introduce is tiering/deduplication to reduce a large amount of data stored in the distributed storage system. This method enables global deduplication with tiering by eliminating deduplication metadata efficiently. Second, I will describe the performance limitation current distributed storage faces and discuss ideas to solve the limitation. Lightweight store and replication offloading will be introduced to improve performance on NVMe-based storage cluster.
  • Myungwon Oh
  • Myungwon Oh / S/W Development Team, Memory Business Samsung Electronics
  • Myoungwon has been contributed storage features such as tier/dedup and performance to Ceph which is one of the most popular open-source distributed storage system as a committer. He is currently working for Samsung electronics. His interest is how the distributed storage system will be transformed by storage evolution.

Feel free to leave comments/curiosity/opinions about the session.

* Comments (including swearwords and slander) that do not match the purpose of the bulletin are deleted by the administrator without notice.

* Comments can only be uploaded while the sessions are ON–AIR and updated every minutes.

0/100 characters WRITE