Paper Title: Implementing Hot Plug in NVMe Storage Systems

Paper Abstract: NVMe is rapidly becoming the high-performance storage interface of choice in data centers. So designers must come up with ways to implement hot-plugging in this new environment. They must identify the impact of host/device timeout on the operating system/BIOS, put mechanisms in place to protect uncorrectable fatal errors from propagating to the host, and update the Linux kernel/NVMe drivers to address all-1s-completion and optimized advanced error reporting (AER) handling. There are also mechanical challenges involved in slowly adding and removing devices from a PCIe topology, and connectors may impact system reliability.

Paper Author: Wesley Yung, Principal Engineer, Applications, Microsemi

Author Bio: Wesley is the Applications Hardware Engineering Lead for Microsemi's PCIe switching products. He currently focuses on switching solutions for rack disaggregation, virtual machine resource sharing, and NVMe over fabric acceleration. Wesley has been providing pre-sales, marketing, architectural, signal integrity, and applications leadership for PCIe switches, SAS/SATA expanders, and Fiber Channel Controllers/Switches for over 15 years. He works with standards organizations including T10, PCI-SIG, and NVMe, and recently did a joint presentation with Facebook on their Lightning NVMe JBOF at the OCP US Summit. He earned a Bachelor’s in Engineering at the University of Victoria (Canada).