With the launch of ONTAP® 9.4, NetApp has introduced the first NVMe over Fabrics (NVMe-oF) implementation of NVMe over Fibre Channel (NVMe/FC). In initial testing across different workloads (a joint effort with Broadcom’s Brocade and Emulex divisions), we’ve seen more than 50% higher IOPS and up to 34% lower latency than with SCSI FCP.

 

Given the variety of different NVMe-oF transports, why did we choose to start with Fibre Channel as our first NVMe fabric implementation in ONTAP? Here are five great reasons:

  • Almost all high-performance latency-sensitive block workloads are running on FCP today.
  • Most performance-focused workloads already use either Gen 5 (16Gbps) or newer Gen 6 (32Gbps) switches in their fabrics.
  • FCP and NVMe/FC can use the same physical components to transport FCP and NVMe/FC concurrently.
  • Many NetApp customers already own all the hardware they need and can enable NVMe/FC with a simple software upgrade to ONTAP 9.4.
  • In contrast, few of the 25/50/100Gbps Ethernet switches required for alternative NVMe-oF architectures have been deployed in most data centers.

Because NVMe/FC and Fibre Channel networks use the same underlying transport protocol (FCP), they can use common hardware components. It’s even possible to use the same HBA, switch, cables, and ONTAP target port to communicate with both protocols at the same time. The ability to use either protocol by itself or both at the same time on the same hardware makes transitioning from FCP to NVMe/FC both simple and seamless.

 

This also means that an organization can provision an application using NVMe namespaces (the equivalent of an FCP LUN) while other applications are using FCP LUNs in the same physical path at the same time. Additionally, because NVMe/FC uses the same FC switch infrastructure, organizations already have the technical skill sets to support both protocols with almost no retraining.

The Industry’s First Enterprise End-to-End NVMe Solution

In addition to the introduction of NVMe/FC with ONTAP 9.4, NetApp also announced the new AFF A800 all-flash array. The A800, with its NVMe solid state drives, will be the industry’s first end-to-end NVMe offering that is both announced and shipping. The performance improvements of NVMe/FC and back-end NVMe drives are additive, which means that end-to-end NVMe will enable the most efficient and highest-performing storage area networks in the industry.

Which Workloads Will Benefit from NVMe Performance?

NVMe technology can significantly boost storage system performance. Why does that matter? By improving block storage performance, NetApp customers can accelerate many existing applications while enabling new workloads for the first time. Some of the workloads that are expected to benefit from NVMe/FC and end-to-end NVMe include:

  • Artificial intelligence and machine learning. The high-performance and ultralow-latency characteristics of NVMe can accelerate I/O as part of a deep learning data pipeline.
  • Real-time analytics. High-performance NVMe storage can be paired with big data analytics to generate new insights for a variety of data types.
  • Internet of Things (IoT). The performance, scalability, and reduction in operational latency make NVMe an obvious choice for processing the massive amounts of data generated by IoT devices.
  • Existing mission-critical applications. In addition to new computing initiatives, NVMe also improves the performance of existing enterprise applications, including Oracle, Microsoft SQL Server, SAP HANA, and NoSQL databases such as MongoDB.

NVMe Information and Resources

To learn more, check out these resources:

Michael Peppers

Michael Peppers is a technical marketing engineer (TME) with a focus on SAN, FlexArray/V-Series, QoS, FLI, and NVMe in ONTAP. His TME mission is to broaden and deepen the SAN, FlexArray & related knowledge of customers, partners, and coworkers, and increase the scope of NetApp's SAN & FlexArray solutions. Prior to becoming a TME Mike tested ONTAP SAN Interoperability in engineering quality assurance, before that, was an Escalations Engineer in the NetApp Technical Support Center. Before joining NetApp in 2005, Mike was a Networking and Systems administrator for several of US-based corporations.