PACKET SWITCHING

One only need examine history over the past couple of hundred years to see that older communications models and methods bear resemblance to some of the current crop of fast growing methods. For example it’s not difficult to see the similarity between smoke signals and telegraph messages. If one takes the smoke signal model and imagines a sender sending a message to a receiver and the receiver repeating the message to another smoke sender and so on, it’s easy to see the resemblance to packet forwarding characteristics of the Internet protocol or email.

If you have any interest at all in the Internet and have done any reading on the subject, you’re aware that it’s based on packet switching. Packet switching depends on some basic functional elements including transmission links, and a switching engine called a router.

There are a lot of similarities between cell-based switching and packet-based switching, and there are some differences. Packet switching is simply making decisions about where to send the packet at hand. Packets are like cells in the sense that they must be opened, intelligence found about where they are headed and where they have been, and then switched and/or routed. The instructions are just inside the packet with a few other tidbits of information.

One of the fundamental differences between packet switching and cell switching is at the heart of most of the ambiguity and hand-wringing that occurs when considering the routing-switching—layer 2-layer 3 solution. It’s really quite simple. Cells, PDH streams, and PPP (HDLC) are layer 2 functions. What separates these techniques from packet techniques is a time base. Packets, at least IP datagrams that make up user datagram protocol and Telecommunications protocol over Internet packets, have no reference or relationship to any clocking, timing, or basic synchronizing intelligence. They are just out there somewhere in the Ether. T1/E1, PPP/HDLC, ATM, and Ethernet all have clocking and synchronizing information embedded in the stream.

Packet switching is nothing more than switching and/or routing at packet borders, or between packets after the details in the packet header have been opened and read. Only after the entity has been opened and read can it be routed or switched to a second port. Many times the mail system is used as a metaphor for packet switching. It’s a pretty good metaphor, but with some subtle differences. First, the packet entity must be opened and read. It does not have an outside and an inside unless the payload has been encrypted, or otherwise sealed and secured in some way. One of the fundamental flaws in the Internet everyone knows about and experiences every day is simple courtesy and security. In addition to the payload and addressing information, there are other significant details inside the packet entity exposed for any and everyone to see and do with as they please. These other details have to do with all kinds of fun things that can muck up the overall machinery such as administrative control of the routing machines.

A view that says the Internet has evolved from prior well-known methods and technology wouldn’t be difficult to contend, but would likely be more difficult to defend. Many modern IP network designers seem blissfully unaware that the Internet is critically dependent on an underlying transmission infrastructure they simply refer to as the network layer. Very few of them have a clue about the importance of network clocking and timing. Many think packet over SONET/SDH isn’t a big deal because it’s done all the time (over PDH), which for the most part goes over SONET/SDH any way. Very few realize that HDLC, or PPP framing, is as rigid and fixed as T1, E1, T3, or E3. A few understand the details of packet-over SONET/SDH. The ones that do understand this fundamental know that Internet architecture includes layer 1 and layer 2 and is not, as the rest of their esteemed colleagues contend, self-healing.

Throw in all the mumbo jumbo about connection-oriented and connectionless protocols and mumble solution in between every fifth use of the word router or whatever else can be thought of, but the basics remain the same. That is, something on the premises, or at the network access point, contains or establishes intelligence that tells the network how to set up a connection between two or more points, and thereby transmit and receive information through the network. Anyone can play around with semantics all day about dumb terminals and smart networks, or at the other extreme, they call intelligent terminals and dumb networks. At the end of the day, what’s important is effective and efficient use of limited resources.

If confusion reigns, stop and ask a couple basic questions: ‘‘Is it circuit, cell, or packet?’’ ‘‘What is being shared?’’ Is it time, bandwidth, or both? What are the interface, bitrate, and active protocols on the facility? What is supposed to be done with it? What did the customer ask for? What is being delivered? Is it broke? With a little patience and perseverance, confusion will soon stop raining, the clouds will pass, and matters will clear up as you climb up or down the stack of bits and bytes.