Data Center Fabric Architectures

This section describes the structure of the webinar materials, lists other webinars you should watch to master data center fabric design, and addresses fundamental questions like:

• What is a data center fabric?
• How do I start designing one?
• Who are the major players?
• Does merchant silicon matter?
• How do I choose a vendor and/or a fabric solution?

What requirements should you consider when designing a data center fabric? This section lists a number of them and explains why they might be relevant to your design.

Regardless of what the vendors love to tell you, every data center fabric solution implements one of common architectures, ranging from completely distributed to completely centralized.

Not surprisingly, each of these architectures has typical benefits and drawbacks; you’ll discover them in this section.

In the introductory part of 400GbE session Lukas Krattiger covered the "big picture" considerations valid for any data center fabric design:

• Typical fabric designs and business drivers
• Leaf and spine role selection
• Buffer sizes and platform considerations

Some of the technologies used in data center fabrics are standardized. Others are not, but they’re still used by most vendors in one form or another.

This section describes some technologies you’ll find in most data center fabric solutions, including Multi-Chassis Link Aggregation (MLAG), Unified Forwarding Tables (UFT), chassis architectures, ASIC buffering, and 400 Gigabit Ethernet.

This part of the webinar contains hardware- and software overview of solutions offered by major networking vendors. It was last updated in 2019 -- hardware information is probably obsolete by now, but you might still find interesting descriptions of older software features. Just keep in mind that whenever a video says "vendor X cannot do Y", they probably fixed that by now (but do check whether they did nonetheless).

This section contains information on Arista’s data center switches and software features available in Arista EOS.

This section covers Cisco’s Nexus series of data center switches and their Nexus-OS operating system.

This section describes Cumulus Linux, an open-source Linux-based network operating system running on whitebox switches as well as switches manufactured by Dell and HPE.

Cumulus was acquired by Mellanox, which was acquired by NVIDIA. Cumulus Linux is currently (late 2023) in release 5 which uses new configuration mechanism -- NVUE (NVIDIA User Experience).

This section covers Juniper’s QFX- and EX-series of data center switches and the data center features of Junos.

These vendors were acquired (Avaya), sliced up and sold for parts (Brocade), or gave up (HP Enterprise). The only one left standing in this category is Dell.

This section contains information on Avava’s (now Extreme) data center switches and their SPB fabric solution as of summer 2016.

This section covers Brocade’s VDX series of data center switches and their NOS operating system. It was last updated in September 2017 and includes hardware descripton of SLX switches (software documentation wasn't available yet).

This section contains a brief introduction Dell data center switches and FTOS operating system. It was last updated in 2015.

In the meantime, Dell focused on BriteBox switches, although it still has its own network operating system (OS10) that is not covered in this webinar.

This section describes Comware-based data center switches sold by HPE. It was last updated in 2015.

This section contains one-off presentations describing minor data center switching vendors. Most of the either gave up or got acquired and disappeared.