Storage and Computing Acceleration for Big Data Analytics

NVRAMOS'16 day2 Tech session 1-2

Posted by Seongjin Lee on October 21, 2016 |

Mookyung Jung (SK Telecom)

big-data 에서 cpu bottleneck으로 storage I/O성능 다 쓰질 못함 dpa - data processing acceleration cpu bottleneck문제 큼

I/O accelerator

  • nvcache
  • key-value interface

Data processing acceleerator data relocation

  • compression : 30% perf down 문제
  • deduplication : 30% perf down 문제 computing
  • sql/filter/table pivot - sw&hw 최적화 중
  • machine learning / dnn

comments powered by Disqus