How Can You Differentiate Between Routing and Switching?
by Howard C. Berkowitz

In this article, Certification Zone Technical Director Howard Berkowitz attempts to answer the following questions: What is IP Switching and how do you differentiate between routing and switching? Can it be defined in terms of L2 and L3 in the OSI Model?

Howard's Response: Part of the problem is that the basic OSI model and its definition of L2 and L3 is obsolete. This isn't a criticism of the OSI people, because there are documents, such as "Internal Organization of the Network Layer", that update the definition and make it more consistent with today's technology. There are equivalent advances in the Internet architecture.

Unfortunately, these updates, while they certainly are used and even originated by Cisco engineers, haven't appeared in the Cisco training and certification material.

To answer the specific question, IP switching was the first MPLS-like commercial product, developed by Ipsilon. It ran only over ATM. Ipsilon was not a commercial success and eventually was acquired, IIRC, by Nokia.

MPLS was a multivendor IETF effort that, frankly, was a response by a very scared industry to what a cleaned-up IP switching could do. It merged and extended several proprietary technologies, including Cisco's tag switching, as well as protocols from IBM, Lucent, and others.

Several years after the IETF MPLS work started, they generalized it and created what was called the "sub-IP" area, which dealt with intelligent protocols below IP/network layer such as MPLS, certain ATM derivatives, optical networking, etc. See

http://www.ietf.org/html.charters/wg-dir.html#Sub-IP%20Area

There is now a Generalized MPLS architecture (GMPLS), which goes beyond the packet orientation of MPLS and includes additional, non-packet transmission systems, such as optical/wavelength/lambda switching/routing, time-division multiplexing like SONET/SDH, and physical port-by-port routes as, for example, seen in digital crossconnects.

To answer your question about routing and switching, this is an area of confusing terminology, a good deal of which is pure marketing handwaving based on a perception that "switching" is somehow faster than "routing". This is not a technically useful definition, especially considering the way any modern high-performance layer 3 relay (avoiding the words "router" or "L3 switch") is actually built.

These terms do get used in some Cisco materials in a more useful way, where "routing" is the process of building the routing table (I'm simplifying here) and "switching" is the process of packet forwarding. It's more correct to speak, instead, of the "control plane" and the "forwarding plane". Inside a high-end router, different hardware and software handle these two functions. There is much discussion of this in the Groupstudy archives.

Now that I've talked about control and forwarding, let me return to GMPLS. GMPLS cooperates with the IP control plane/routing to become aware of a network topology, and the GMPLS control plane uses protocols such as RSVP-TE and LDP to set up a forwarding plane for packet and non-packet information.