Ethernet Storage Fabric (ESF) delivers performance and efficiency for scale-out storage and hyperconverged infrastructures. It leverages the speed, flexibility, and cost efficiency of Ethernet to provide the foundation for the fastest and most efficient storage networking infrastructure.
ESF runs on purpose-built switches which are optimized to deliver the highest levels of performance, lowest latency and zero packet loss, with unique form factor and storage aware features. Other capabilities of ESF include simultaneous handling of compute and storage traffic, future proofed with support for the NVMe over fabric protocol, support for file, block, and object storage, and it is best suited for scale-out storage and Hyperconverged infrastructures.
This section describes NVIDIA Ethernet Storage Fabric solution, its use cases, implementation and monitoring and debugging capabilities.
The most common deployment of ESF is a single rack of 6-18 servers, or in the case of HCI 6-18 appliances. The servers/appliances are connected in high availability architecture, utilizing MLAG, to two ToR SN2100/SN2010 half ``19 width Spectrum switches, enabling high availability in a single rack unit.
We will start with the setup/topology overview, followed by its Bill of Material and connectivity guidelines.
The following sections will describe the various ESF deployment manners available for the user:
- CLI based configuration done one-by-one on all switches
- Automation based configuration using Ansible
- Using NEO as the management system