Performance analysis of I/O intensive workloads (such as SPDK-based applications) appears quite challenging, since system performance depends not only on how effectively code is executed on CPU cores, but also on CPU/IO devices interaction efficiency.
In this talk, you will learn about:
- oC-level activities induced by PCIe traffic processing and possible pitfalls
- Best known methods to achieve best performance on Intel® Xeon® Scalable processors
- How to detect these performance issues using Intel® VTune™ Profiler’s Input and Output analysis, which reveals system topology, hardware resources utilization (PCIe, Intel UPI and DRAM), and Intel® Data Direct I/O (Intel® DDIO) traffic processing efficiency