Chapter 7. IP Fabrics (Clos)
Everywhere you look in the networking world you see something about IP Fabrics or Clos networks. Something is brewing, but what is it? Whatâs driving the need for IP Fabrics? If an IP Fabric is the answer, then what is the problem weâre trying to solve?
Many over-the-top (OTT) and software-as-a-service (SaaS) companies have been building large IP Fabrics for a long time, but have rarely received any attention for having done so. Such companies generally have no need for compute virtualization and write their applications in such a way that high availability is built in to the application. With intelligent applications and no compute virtualization, it makes a lot of sense to build an IP Fabric using nothing but Layer 3 protocols. Layer 2 has traditionally been a point of weakness in data centers with respect to scale and high availability. Itâs a difficult problem to solve when you have to flood traffic across a large set of devices and prevent loops on Ethernet frames that donât natively have a time-to-live field.
If companies have been building large IP Fabrics for a long time, why is it that only recently IP Fabrics have been receiving a lot of attention? The answer is because of overlay networking in the data center. The problem being solved is twofold: first, network agility, and second, simplifying the network. Overlay networking combined with IP Fabrics in the data center is an interesting way of providing both agility and simplifying ...