Open-E JovianDSS Up32: Revolutionizing Storage Performance with NVMe over Fabrics Initiator and Multipath I/O

The landscape of data storage has been profoundly reshaped by the advent of NVMe, an interface specifically optimized for SSDs and designed for high-speed media. NVMe distinguishes itself by supporting high parallelism, boasting up to 64,000 queues, each capable of handling up to 64,000 commands, all operating efficiently on the same processor core. This architecture was a significant leap from older SCSI standards, which were not designed to leverage the full speed of modern flash storage.

NVMe-oF: The Next Generation of Data Storage Connectivity

The Evolution of Data Storage: From NVMe to NVMe-oF

The evolution from direct-attached NVMe to NVMe-oF (NVMe over Fabrics) represents the next frontier in data storage connectivity.

NVMe-oF is a protocol specification engineered to connect hosts to data storage across a network fabric, utilizing the high-performance NVMe protocol. Unlike local NVMe, which maps commands to shared memory over a PCIe interface, NVMe-oF employs a message-based model for communication between a host and a target storage device. This architectural shift is driven by the imperative to eliminate bottlenecks inherent in traditional SCSI-based protocols, such as iSCSI or SAS , which were not initially conceived for the blistering speeds of contemporary SSDs.

The Benefits of NVMe-oF Utilization

NVMe-oF allows connecting to and accessing NVMe-based storage arrays over both TCP and RDMA, which, particularly when coupled with RDMA, significantly reduces overhead and latency.

RDMA facilitates memory-to-memory transfers between two computers without requiring intervention from the operating system or the processor, resulting in latencies typically measured in microseconds. The combination of NVMe-oF and RDMA enables the full potential of NVMe SSDs to be realized across a network, bridging the gap between direct-attached storage and SANs.

Boosting Performance and Availability with Multipathing (MPIO) in Open-E JovianDSS

Open-E JovianDSS Up32 introduces NVMe-oF Initiator functionality, enabling the system to connect to NVMe-oF targets. A critical component of this integration is the inclusion of MPIO (Multipath I/O) functionality. Multipathing establishes multiple simultaneous paths between the NVMe host initiator (Open-E JovianDSS) and the data storage target. This capability offers substantial advantages:

• Performance Enhancement: By distributing the I/O load across multiple paths, MPIO significantly increases aggregate throughput, ensuring that the high performance of NVMe-oF is fully utilized.

• High Availability and Redundancy: MPIO provides robust failover mechanisms. If one path experiences a failure, I/O operations are automatically rerouted through an alternative path, ensuring continuous operation and bolstering data resilience.

• Scalability: The architecture supports the simultaneous sending and receiving of commands from numerous hosts and storage subsystems, allowing for extensive scaling of data storage environments.

How NVMe-oF with Multipathing Transforms Data Center Infrastructures

The incorporation of NVMe-oF with multipathing into Open-E JovianDSS marks a significant progression towards constructing highly resilient, high-performance shared data storage infrastructures.

These infrastructures are designed to support the most demanding enterprise workloads, including large-scale virtualization deployments, complex databases, and intensive analytics platforms. This advancement facilitates deeper consolidation within the data center, effectively eliminating data storage silos and enabling the leverage of a single, highly efficient shared storage infrastructure. Open-E‘s adoption of NVMe-oF is also indicative of a broader industry movement towards disaggregated data storage architectures. In such models, compute and data storage resources can be scaled independently and accessed over high-speed networks, thereby maximizing flexibility and overall resource utilization.

For enterprise workloads, the practical benefits are profound. The solution empowers organizations to support applications that demand high throughput and exceptionally low latencies. It also contributes to reducing the length of the operating system storage stacks on the server side and offers a faster overall solution by facilitating the transition from traditional SAS/SATA drives to high-speed NVMe SSDs.

NVMe-oF vs. Traditional Storage Protocols

To further illustrate the advantages, a comparison of NVMe-oF with traditional data storage protocols is presented below:

This table highlights the architectural and performance advantages of NVMe-oF over older technologies, providing a clear comparison of how it addresses the demands for low latency, high throughput, and increased parallelism critical for modern enterprise applications.

Explore More Open-E JovianDSS Up32 Features

Beyond the powerful capabilities of NVMe-oF and MPIO, Open-E JovianDSS Up32 is a comprehensive leap forward in data storage technologies. This new version introduces VMware VAAI support for NFS, a new Storage Pool Initialization function, and configurable power button settings. These features, along with robust backup options, enable you to build a truly resilient high-availability data storage cluster with the redundancy and performance required for modern enterprise workloads. It also features a fully updated Linux kernel, major driver updates for network and storage controllers, and enhanced ZFS performance with improved TRIM support. If you want to know more about this, read the article about all Open-E JovianDSS Up32 features.