Leaf Spine Architecture in DC

Let’s look at a more recent switch architecture type known as leaf/Spine. It is becoming apparent that fiber connectivity in data centers is getting more complex and that traditional methods for fiber connectivity management may not be sufficient. 

Data center design is undergoing a transformation:  A traditional data center, with core/aggregation and access switches, focuses on north-south traffic between application components residing on multiple servers and therefore it fails to adequately address latency, which is critical in many of today’s applications.

A more direct path for server-to-server communication is achieved with “spine-leaf” architecture. In such a design there is a high density of fiber connections between leaf and spine switches which facilitates east-west traffic in support of “any-to-any” network connectivity.

Another trend is a need for upgradable cabling infrastructure. Upgradeability of infrastructure from 10G to 40G and beyond, necessitates reconfiguration of fiber cabling infrastructure with a need for simplifying the upgrade path making it as seamless as possible to minimize risk.

Higher bandwidth also places requirements for providing fiber infrastructure with flexibility of using array connectivity breakouts in support of a variety of applications and to facilitate transition between parallel and serial optics.

As for network port density with space at a premium in the data center, increased density of fiber ports on equipment as well as on fiber shelves must be supported adding to usability challenges associated with managing fiber connectivity.