Flash Memory Summit (FMS) 2022 – NVM Express Presentation Track

NVM Express® (NVMe®) has become synonymous with high-performance storage with widespread adoption in client, cloud, and enterprise applications. The NVMe 2.0 family of specifications, released in June 2021, allow for faster and simpler development of NVMe solutions to support increasingly diverse environments, now including Hard Disk Drives (HDDs). The extensibility of the specifications encourages the development of independent command sets like Zoned Namespaces (ZNS) and Key Value (KV) while enabling support for the various underlying transport protocols common to NVMe and NVMe over Fabrics (NVMe-oF™) technologies. This presentation provides an overview of the latest NVMe technologies, summarizes the NVMe standards roadmap, and describes new NVMe standardization initiatives.

To facilitate the development of the refactored NVMe 2.0 family of specifications, the public release of several NVMe, NVMe-oF and NVMe Management Interface (NVMe-MI™) technical proposals (TPs) were delayed but were recently published publicly. In this presentation, attendees will learn about those new TP’s contained in the NVMe 2.0 specifications and new TP’s not contained in contained in the NVMe 2.0 specifications since its publishing. Included in the presentation is Domains and Partitions, Copy Command, Protection Information Enhancement, Disperse Namespaces, Centralized Discovery Controllers, Zone Random Write Area and more. Join us to find out how these new features will aid NVMe technology development and propel us into the future of storage.

With the release of the NVMe 2.0 library, command sets have been separated out into individual standards. With this refactoring, NVM Express is enabling implementers to better utilize the command sets relevant to their specific use cases. As part of this refactoring, the NVM Express Technical Working Group released the NVM command set, NVMe Key Value (KV) command set, and NVMe Zoned Namespace (ZNS) command sets as separate specifications. Attendees will learn about these new specifications and how NVMe-KV and NVMe Zoned Namespaces offer crucial performance benefits such as reduced overprovisioning, increased transactions per second and improved control of I/O and data placement.

With the availability of separate NVMe transport specifications (NVMe/PCIe Transport specification, NVMe/RDMA Transport specification and the NVMe/TCP Transport specification) NVM Express can isolate and independently evolve NVMe-oF transports for emerging technologies and use cases. In this presentation, NVMe technology experts from HPE and Lighbits Labs will explain the changes to the NVMe-oF specification, what’s new in the transport specifications and the unique capabilities and benefits of deploying NVMe technology with each transport.

A review of the latest NVMe device types entering the market along with a future view of a consolidated storage architecture based on an NVMe deployment model. This top-level architecture and deployment model will show how multiple facets of the NVMe protocol supports scale and direct attached storage for multiple media types and accelerators. We’ll take an in depth look at the newly supported rotational media (HDD)type and how this provides the overall opportunity for a Total Cost of Ownership (TCO) savings supported by a consolidated NVMe architecture.

NVMe over IP-Based fabrics, such as NVMe/TCP and NVMe/RDMA, has the potential to provide significant benefits in application environments ranging from the Edge to the Data Center. However, before we can fully unlock the potential of NVMe over IP-Based fabrics, we first need to overcome its discovery problem. This discovery problem, specific to IP based fabrics, can result in the need for Host administrators to explicitly configure each Host to access each of the NVM subsystems in their environment. In addition, any time an NVM Subsystem interface is added or removed, the Host administrator may need to explicitly update the configuration of impacted hosts. This process does not scale when more than a few Host and NVM subsystem interfaces are in use. Also, due to the de-centralized nature of this process, it also adds complexity when trying to use NVMe over IP-Based fabrics in environments that require a high-degrees of automation. For these and other reasons, several companies have been collaborating on innovations that simplify and automate the discovery process used with NVMe IP based SANs.

This session provides an overview of the NVMe software drivers, including new updates and what is supported. The presenters will describe the vSphere support for NVMe and NVMe-oF technologies with an overview of the ESXi driver eco-system followed by an update of the various NVMe features support especially pertinent to NVMe-oF including VMware support for some of the NVMe-oF control plane features. Updates will be provided on UEFI Device Drivers including NVMe related updates in UEFI Platform Firmware, Built-in Diagnostics, where and how the UEFI drivers fit in the NVMe Software Driver Eco-System and other resources to enable developers and NVMe device, software, and platform vendors. Finally, this session will provide updates to the Windows in-box NVMe drive.

We will be speaking about testing the latest and greatest NVMe and NVMe-oF features from both 1.4 and 2.0, and how it is changing with the refactoring of the NVMe technical specifications. The UNH-IOL has adapted the existing NVMe Conformance and Interoperability test plans and IOL INTERACT test tools to align with the 2.0 refactored technical specifications.

In NVMe-oF environments, improving high speed data transfer into multicore systems is a well-known challenge. Methods like network adapter offloads and performance tuning are effective practices in this domain. A lot of attention has been focused on how to set these tuning parameters and which offloads to select for higher performance. One of them is optimizing CPU utilization, which is most of the time platform dependent. By using smart mapping between software and hardware threads (affinity), better resource utilization can be achieved. As the NVMe-oF layer is the main software layer that is sensitive to hardware resources in terms of sharing and availability, this will affect performance under both TCP and RDMA transport types. This session will present a software-based mechanism that aims to optimize results and achieve performance scalability through automatic CPU core affinity settings. This innovative approach challenges the common use case of systems using CPU masks for each NVMe-oF portal, risking poor resource utilization. This session will examine the use of smart CPU Core affinity, suggest best practices, discuss NVMe-oF transport limitations and more.

With the increased availability and awareness of NVMe products, more enterprises are making the transition to NVMe architecture and it is expected it to be the dominant storage interface for enterprise going forward. Join our panel discussion to learn how server and storage vendors are helping make this transition possible and how the release of the NVMe 2.0 specifications is shaping the future of enterprise storage markets.

SSDs struggle with write amplification, performance, QOS and garbage collection. Flexible Direct Placement Mode represents a significant innovation from Google and Meta addressing these issues which enables significant TCO benefits compared to today’s SSDs.

Hyperscale storage is a challenging and growing part of the storage eco-system. NVMe is an important technology in meeting the cloud needs. This talk will discuss the needs, challenges and how NVMe is deployed at scale to meet hyperscale needs.

Hyperscale data centers are software-defined. From network resource to the FPGA and NPU accelerators in servers to the storage underlying it all, software-defined provides significant benefits to run and manage the data center. A Linux Foundation project, Software-Enabled Flash technology applies software-defined capabilities to unlock the power of NVMe flash storage, allowing hyperscale developers to manage NVMe storage down to the die. See how the open source API allows developers to assign application and use case profiles to NVMe based hardware modules to optimize performance, latency, and maximize the life of flash.

Did you ever imagine that NVMe could be used in a memory card the size of fingernail? The world’s most popular removable memory cards, SD and microSD, added the PCIe and NVMe interfaces into the SD Express and microSD Express specifications. With this adoption, NVMe is everywhere and across all parts of the storage ecosystem from data centers to removable memory cards. This talk will describe the SD Express specifications (SD7.x and SD8), upcoming new features, how to easily implement SD Express in host devices, plus describe the world of opportunities created with removable cards which support the NVMe storage interface.

CompactExpress, SD Express, XFM, E1.S, E1.L, E3, SFF-TA-1001, M.2 Adapter, M.2 BGA, Automotive SSD BGA, SFF Connector Module (U.2), CEM, oh my! Who defines each form factor, What are the lanes, speeds, & common capacities (as of mid-2022), Where are they commonly found.