Facebook on Tuesday described a disaggregated network system, called the Fabric Aggregator, which it’s using to scale network capacity between data center fabrics in its regions.
Using simple and open building blocks — such as the Wedge 100, Facebook’s top-of-rack network switch, and FBOSS, a Linux-based operating system — the design of the Fabric Aggregator provides the flexibility needed to quickly adjust capacity.
The company needs the flexibility as builds its twelfth data center and expands its Papillion, Nebraska location from two buildings to six, Facebook explained in a blog post. Its data center footprint is growing to accommodate more immerserve content, such as live video, 360 photos and virtual reality.
In order to provide more flexibility and operational simplicity, Facebook developed a cabling assembly unit for the Fabric Aggregator, which emulates the functionality that a backplane would offer in classic chassis-based devices. In other words, it provides interconnection between the building blocks within the Fabric Aggregator.
“With this approach, we can scale the capacity, interchange the building blocks and also change the cable assembly quickly as our requirements change,” the Facebook blog post explained.
Facebook designed four different cabling configurations and is sharing them via the Open Compute Project.
Meanwhile, ach Fabric Aggregator node — a unit of bandwidth replicated to meet the overall demand of an aggregation tier — implements a two-layer cross-connect architecture. The downstream layer is responsible for switching regional traffic (east/west) while the upstream layer is responsible for switching traffic to and from other regions (north/south).
“Separating the solution into two distinct layers allow us to grow the east/west and north/south capacities independently by adding more subswitches as traffic demands change,” the blog post said.
The building block approach, which enables operation at either the subswitch or node level, offers flexibility and security, Facebook explained. “The Fabric Aggregator layer can suffer many simultaneous failures without compromising the overall performance of the network.”