Photo by Mathew Browne on Unsplash

NVMe-oF gears up to replace iSCSI

Changing of the Guard

Article from ADMIN 84/2024

By Martin Loschwitz

NVMe over Fabrics promises high-redundancy centralized storage with simple maintenance and management and almost twice the speed of its antiquated iSCSI predecessor.

Centralized storage in the form of network-attached storage (NAS) or a storage area network (SAN) are implicitly or explicitly set in stone in the data center. Wherever you look, you will likely find a device whose sole task it is to provide persistent storage to the other systems.

From the administrator's point of view, it makes sense to handle central tasks such as redundancy and backup with a single point of administration. The alternative would be uncontrolled local growth with complicated backup and redundancy mechanisms. Of course, local storage is occasionally inevitable (e.g., in cloud setups where virtual machines require extremely fast storage). However, cloud providers tend to assume that the customer's application takes care of data redundancy and regular backups itself. For almost all other scenarios, centralized storage is the ideal solution for system administration.

Like any technology, centralized storage is subject to the whims of time. For what felt like an eternity, the Fibre Channel (FC) protocol was the gold standard, and every virtualization server in the setup needed its own host bus adapter (HBA) to connect it to the FC network. The manufacturers of centralized storage have probably made a fortune over the years simply by selling Fibre Channel hardware (Figure 1), without even taking into account that storage appliances are not exactly kind on the customer's wallet.

...

Use Express-Checkout link below to read the full article (PDF).