Mike, a great article from IBM et all, HLIT works with them over in Israel. Anyway it addresses my contention at to how to implement the future technologies for last mile.
bcr.com
In the SONET multiplexing structure, high-speed signals can be created in two ways. As shown in Figure 2, a number of discrete STS-1 channels, each with its own payload and overhead, can be multiplexed together. SONET also defines concatenated formats where the payloads from several STS-1 equivalents are combined to create one higher-capacity channel. The concatenated formats are identified by the suffix "c," and the STS-3c and STS-12c formats are currently defined as transport options for ATM and other network services.
Now, in addition to these two possibilities, DWDM provides an additional multiplexing option: Several optical signals can be combined onto the same fiber path. Because WDM operates at the optical range, it couldn't care less about the digital multiplexing format that is being used. That means SONET and non-SONET fiber systems, as well as analog fiber systems used for cable television, can all be combined onto the same fiber
New Switching Options
Clearly what we are getting here is a whole new set of options for how to switch, multiplex and transmit information in a high-capacity network (Figure 4, p. 26). By itself, SONET was a major step in building flexibility and functionality into a fiber network. The SONET terminal is the main serving device for lower-capacity channels; SONET add/drop muxes allow the carrier to reconfigure channels at Layer 1 (i.e., digital cross connect), using commands sent in part of the SONET overhead. If the system used virtual tributary mapping, the SONET device could cross-connect individual DS1 channels. If the interface to the SONET environment were an asynchronous DS3 (i.e., the output of an M13 multiplexer), the SONET network could only cross-connect the entire DS3.
The development of DWDM and high-capacity ATM and Layer 3 switches will provide new options for how switched or dedicated services can be delivered to users. With DWDM, each independent fiber transmitter is assigned a different wavelength, which means we can switch different wavelengths onto different paths.
The ability to switch those different wavelengths is called lambda switching or photonic switching. (If SONET is doing Layer 1 switching, lambda switching must be Layer 0.5!) The idea in a photonic switch would be to use a switching network to break out the wavelengths of the various fiber inputs and recombine them into new outputs.
The first step toward photonic switching--optical cross connects and add/drop multiplexers--is no longer a distant prospect. Ciena Corp. (www.ciena.com) and Harmonic Lightwaves are developing products for rearranging optical channels. Lucent (www.lucent.com) introduced a preliminary 32-line optical cross connect system last year. Others, like Astarte Fiber Networks (www.starswitch.com) and Cambrian Systems (www.cambriansys.com, a division of Newbridge) are also planning products in this area. Chorum Technologies (www.chorumtech.com) and Tellium (www.tellium.com) are developing optical add/drop multiplexers for WDM fiber networks.
Other suppliers, like Lightwave Microsystems (www.lightwavemicro.com) and Optical Micromachines (San Diego) are building components for switching optical signals. The preliminary versions of these devices are quite costly, but according to Mr. Levi of Harmonic Lightwaves, photonic switches should be readily available and cost effective in two to five years.
While lambda switching gives us a Layer 0.5 switching capability, ATM switches and Layer 3 switches give us Layer 2 and Layer 3 capabilities. If you have been following recent announcements from established companies like Bay and Cisco, as well as upstarts like Avici Systems and Juniper Networks
(see BCR, May 1998, pp. 56-59), you know that router capacities are exploding.
One of the great surprises in network technology in the past few years is that we can do hardware switching without ATM. While many of the Layer 3 switching devices actually use ATM under IP (i.e., the IP datagram is split up into a series of 48-octet chunks and sent on an ATM virtual circuit), others like Avici are looking at switching IP directly, without depending on ATM. In either case, SONET will define the standard Layer 1 trunk format, and the SONET frame will be filled with either ATM cells or IP datagrams.
If we can build that SONET interface directly into the ATM or Layer 3 switch, it could produce traffic that is wrapped and ready for SONET and connects directly to DWDM equipment. In fact, Ascend Communications and Williams Communications recently announced plans for just such a new network infrastructure.
What all of this means is our switching options are sliding all over the OSI model. Given their capacities, ATM and IP switching devices have been designed to deliver end user services. If we have an ATM or Layer 3 switch with sufficient capacity, we might be able to eliminate digital cross-connects. We are already starting to see this "layer creep" in campus networks, where Layer 3 switches are reaching the same price/performance point as Layer 2 LAN switches. Why bother cross-connecting Layer 1 pipes when it's just as cheap to handle all the information on a cell or datagram basis?
Hiram |
