Data center interconnection applications have become an important and rapidly growing part of the network domain. The huge growth of data promotes the construction of data center parks, especially the construction of super-large data centers. Now, several buildings in a campus must be connected with sufficient bandwidth. How much bandwidth is required to maintain the flow of information between data centers within a single campus? Each data center can transfer up to 200 Tbps of capacity to other data centers today, but more bandwidth will be required in the future.

What drives the huge demand for bandwidth between campus buildings?

      First, the exponential growth of east-west traffic is supported by equipment-to-equipment communication. The second trend has to do with the adoption of flatter network architectures, such as the spinal lobe network or the CLOS network. The goal is to have a large network structure within the campus, which requires a large number of connections between devices.

      Traditionally, data centers are built on a three-tier topology composed of core switches, aggregation switches and access switches. Although mature and widely deployed, the traditional three-tier architecture is no longer able to meet the increasing workload and delay requirements of the environment. In response, today‘s very large data centers are migrating to the ridgeback architecture. In the lobed architecture, the           network is divided into two stages. The ridge phase is used to aggregate packets and route them to their final destination, and the leaf phase is used to connect the host side and load-balanced connections.

Ideally, each leaf switch fans out to each ridge switch to maximize connections between servers, so the network needs a high-decimal ridge core switch. In many environments, large ridge switches are connected to higher-level ridge switches, often called park or polymer-ion ridge switches, to connect all the buildings in the park together. Because of this flatter network architecture and the use of high-precision switches, we expect to see networks become larger, more modular, and more scalable.