利用实际工作负载与人工智能优化NVMe SSD和持久性内存性能.pdf

Using Real World Workloads and Artificial Intelligence to Optimize NVMe SSD and Persistent Memory Performance Eden Kim, CEO Calypso Systems, Inc. Yafei Yang, CEO DapuStor (大普微电子) Data Storage Summit December 3-4, 2019 Beijing, China Agenda Part I: Real World Workloads – Eden Kim, Calypso • What Why Real World Workloads are Important • How Where to Capture Real World Workloads • Reference Workloads: Synthetic Real World • Performance Comparisons: Synthetic Real World • AI Optimization: Replay Tests as AI Training Loops Part II: Performance Tuning with Real World Workloads – Yafei Yang, DapuStor • Challenges for Real World Workload Performance Tuning • SSD IO Path Architecture • Using Calypso IPF IO Capture Replay • Machine Learning Summary: Question Answer Part I: Real World Workloads Eden Kim, Calypso Systems, Inc. Real World Workloads: What are They, Why are They Important? IO Streams that Occur During Application Use Real World Workloads Real World Workloads are Change as they Traverse Constantly Changing the Software/Hardware Combinations of Stack from point to point Queue IO Streams QDs Depths