With the performance of storage technologies (media, protocols, networks, capacities, etc.) increasing more than hundreds of times in the last few years with the introduction of flash-based solid-state drives, SSDs, especially SSDs connected over NVMe. Non-volatile memory express, NVMe, is a storage and data transfer protocol which enables a solid-state drive, SSD, to take full advantage of a high-speed peripheral component interconnect express, PCIe, bus in a computer by connecting directly into it.
As the performance of the storage gets faster, the bottleneck gets pushed out to the network wire speeds and protocols. Though wire and network speed can be improved via new generations of ethernet and InfiniBand speeds, without faster protocol technologies, these new SSDs and NVMe technologies will have their performance bound up inside the server.
NVMe over-Fabrics, NMVe-oF, is a groundbreaking new storage networking technology, that can take full advantage of SSDs performance across an ethernet or an InfiniBand network. NVMe-oF is designed to leverage the performance of NVMe technology across the network using RDMA, remote direct memory access.
RDMA enables data in memory to be transferred between computers and storage devices across a network with little to no CPU hit or usage. NVMe-oF technology enables extremely high bandwidth with low fabric latency, scalability, management, and storage fault isolation. NVMe-oF also improves utilization of the storage, rack space, and power, while sharing storage across multiple servers. Many applications take advantage of shared storage architecture like Oracle RAC. Computing storage can be disconnected from each other or disaggregated.
Truth in IT sees NVMe-oF as a very compelling technology who’s prevalence will only increase as the demand for storage performance needs to keep pace with the growth in storage capacity. As more applications starve for better performance, new architectures like NVMe-oF are critical to minimize costs so they too keep pace. Cost, performance, scalability can all be kept in balance via NVMe-OF so it’s definitely worth exploring.