Collapsed Backbone and Switching

This segment is on collapsing the campus backbone.

I would like to talk to you about some of the campus networks that we have today, the types, the technologies, and some of the limitations that we are experiencing now, some of the limitations that we are going to be experiencing in the near future. I would like to talk about some of the new technologies that are coming down the road and the demands that they are placing on our campus backbone. We are receiving increasing band width demands from our user. We are taking a look down the road at voice, video, and data integration problems . I�d also like to talk about some of the switching technology that we hope is going to get us through this transition. I�d like to talk about what collapsing the backbone is, some of the components that make it up and also some of the different architectures that you can take a look at. And finally, I�d like to touch on a proposed migration path from current architectures to a switched infrastructure.

Campus networking has two major components, intra-building connections and your inter-building or backbone connections. On your intra-building connections, you are looking at shared media Local Area Networks such as ThickEther or ThinEther or a combination of concentrator hubs and repeaters tying together unshielded twisted pair wire plants. How both of these technologies provide Local Area Network services, allow print sharing, access to CD-Rom towers, file services and access to the backbone and Internet. However, both of these technologies are shared band width; that is, if you have 100 users on a Local Area Network, an individual user has to contend with the other 99 for access to the network. What this leads to is high-speed networks providing very slow response. Now this has been adequate in the past because there have been fairly modest demands put on Local Area Networks. But today our networks are congested, and in the future, it's only going to get worse.

On inter-building connections, or the backbone, what you are typically going to see is a backbone network. Now this is a network of networks -- the way that the building Local Area Networks tie in to each other, to central services and also onto the Internet and other wide-area network connections. There are two ways to connect to a backbone network. You can use bridges or routers. A bridge is a much simpler device and is used to connect the same-technology Local Area Networks into the background network. That is, if you have an Ethernet within the departmental LAN, then you can connect to an Ethernet backbone. However, there are limitations to bridges; they do not allow you to tie together token ring technology to Ethernet or to other technologies. A router is a much more sophisticated device, and allows you to connect different technologies together such as Ethernet to token ring or Ethernet to FDDI. And this is necessary on the backbone because the backbone is a network of networks. One of the most important functions of a backbone network is to isolate traffic from intra-building Local Area Networks. That means keeping the traffic that is inside one building on their Local Area Network from interfering with the Local Area Network in another building. Now this follows from what is know as the 80/20 rule; that is, 80% of the traffic on a area local network stays on that Local Area Network, with only 20% of the traffic needing to hit the campus backbone and perhaps on to the Internet.

Now some of the area local network technologies that we are going to be seeing are Ethernet -- and there are several flavors of Ethernet: the original is 10Base5, or what is known as ThickEthernet; 10Base2, or ThinEthernet; 10BaseT; and 10BaseFL. 10Base5 and 10Base2, Thick and Thin Ethernet, are shared media technologies, that is, they actually use a cable to perform the network function. 10BaseT, on the other hand, is a hub-based technology. That means it uses a piece of equipment known as a hub to tie together individual users on to a Local Area Network. 10BaseFL is also a hub-based technology, much like 10BaseT except that the adapters use fiber optic cable to connect to the user or to other hub equipment.

Another technology you may see is called 802.5 or Token Ring. This is used typically in IBM installations and will normally run over a shielded twisted pair. Finally, for your backbone technology, most campuses have migrated to what is known as Fiber Distributed Digital Interface, or FDDI. This a 100-megabit technology that provides high through-put, redundancy, and has been very useful in the past in providing the connectivity that is required between different buildings.

This is an example of what might be on a typical campus today. As you see, you have a central FDDI ring providing connectivity to different devices that would be located in the buildings around your campus. These devices, either bridges or routers, provide connectivity to the Local Area Networks within each building. Now these Local Area Networks can be of several types, as we mentioned before. The purpose of the router or bridge within each building is to provide connectivity between the different Local Area Networks and also connectivity up to the central backbone.

So why is this no longer an acceptable solution? Well, there are new demands being placed on our campus network. We have more users, more traffic all the time. This is simple arithmetic � when you go from 2,000 users to 5,000 users, your traffic goes up. We are seeing more high-band width, multimedia applications. When the web exploded three or four years ago, we saw the ability to pull down easily graphics and now multimedia video and audio tapes. We are now seeing work group and client server software that are putting new demands on the backbone. Users are demanding immediate response, and it has to be response across the entire campus, not just on their Local Area Network. Another new demand that we are seeing on our campus networks is integration of voice and video. Video conferencing places demands not only on band width but on latency. There is also a need for isochronous connections, that is, connections that provide the timing to allow speech and visual data to be delivered smoothly and without stops or gaps. And finally, what we are seeing is that the 80/20 Rule is now becoming the 50/50 Rule with the new Intranet/Internet capabilities. That is, no longer do you see most of the traffic on a Local Area Network destined for another individual site on that Local Area Network. Now you are seeing greater and greater amounts of traffic destined for the backbone or onto the Internet.

So what are the problems with the current network architecture that don�t support the problems that we�ve just seen? Well, for one thing, Ethernet, Token Ring, and FDDI are shared resources, that is, each user is competing with every other user on the Local Area Network or on the background for access into this band width. There is a lack of support for voice and video. Latency has been greatly increased through the use of routers, and there is no provision for timing across current network technology. We are seeing the cost for high-speed Local Area Network technology and for router ports continuing to be higher than the cost for comparable switching technology.

There is no clean migration path from our current architecture to a switch technology such as switched ether or ATM. And finally, what we are seeing is a multiple locations of network devices within the buildings connecting the Local Area Networks to the backbone, making management difficult.

What I would like to do now is compare switching versus routing. Switching is a Layer 2 technology, that is, a data technology that is simpler than routing. This results in reduced latency of connection and increased speed. Routing, on the other hand, is a Layer 3 technology; that is, it works at the network layer. It is still required for multi-technology integration, such as tying an Ethernet Local Area Network to a Token Ring Local Area Network. And it is also still required for wide area network interconnect. With your connection to the Internet, you will have a router in place somewhere. The difference between switching and routing is that routing is much more expensive for the through-put, that is, a 10-megabit Ethernet connection on a switch is much cheaper than 10-megabit Ethernet port on a router.

There are new switches coming down the road that have some routing functionality. Now they won�t allow you to tie an Ethernet into a Token Ring Local Area Network, but they do provide increased addressing capability: that is, they can look at the Layer III or the network information on an IP packet, for example, and provide the routing that is necessary. They are typically lower in cost than comparable routers, and they also have the capability of providing virtual Local Area Networks, or VLANs, which allow protocol or Mac based segmentation. This allows you to segment your users based on the protocol they are using, such as IP for Internet traffic or IPX for Novell traffic, or based on location and their machine identification.

ATM, or Asynchronous Transfer Mode, is one of the new switching technology. It is based on cell switching, using a 53-byte cell. It provides virtual circuits which, in addition to quality of service enhancements, will allow you to pass voice, video, and data and provide an isochronous connection, or a connection with timing built in. ATM has a high-speed capability. Current implementations are from 25 Mbps up to 622 Mbps. There are higher speed connections available beyond 2.4 Gbps and the ATM technology itself does not place bounds on how fast this can go. In the test beds right now, you can see connections going up to 10 Gbps. ATM provides isochronous connections that support voice and video on the campus backbone.

So what is collapsing the campus backbone, anyway? The first step would be replacing the backbone network, that is an FDDI or other backbone network technology with a single device. What this allows you to do is take that 100 Mb FDDI connection and increase its speed to the backplane speed of the device. This is typically measured in the Gb range. The next step would be providing unshared, switched connections to the desktop. What does unshared mean? It means moving from an environment where local area connection provides a shared bit of band width to where a local area connection is simply a logical interconnect between many people with full Ethernet capabilities. It also means eliminating or reducing the departmental bridges and routers. Since you do not have separate technologies that you need to tie together, you do not need separate bridges and routers within the buildings. Instead, you depend on a switching infrastructure to tie your departmental and building Local Area Networks back to a campus infrastructure. And finally, it means creating a structured network design to facilitate migration to newer technologies.

Now what is structured network design? It means providing a connection directly from the campus backbone to the end user with as few interruptions as possible. If you can go from the campus backbone to the end user with a single interruption for a piece of equipment, then you will assure that, as technologies change, you will be able to replace that piece of equipment with the new technology.

So what are some of the advantages of a collapsed backbone? Higher speeds; you aren�t limited by a network technology, you are simply limited by the backplane speed of a central hub. Simplified network configuration; rather than dealing with the needs and requirements of multiple network types and typologies, you have a single standardized configuration which is carried through the entire network. A lower total cost; the cost per port on switching fabric is much less than cost per port for routers. And finally, it provides a central point for the management of your network.

What is a migration path to a collapsed backbone? What we are looking at are several steps you can take to what is in place now to a collapsed backbone architecture without throwing everything that you own now out the window. The first step is to take a look at collapsing your networks around an area router architecture. Second, install central star wiring supporting switch connections within your building. Third, install a fiber star backbone using current technology and equipment. And then finally moving towards an ATM backbone. Let�s take a look at each of these points in a little more detail.

The first step is collapsing around an area router architecture. Now what this means is actually taking the bridges and small routers that are located within your buildings and centralizing them, making a central router the backplaying over which the traffic between different buildings and Local Area Networks travels. Rather than going through multiple devices, such as multiple bridges and multiple small routers, you have a single router performing this function for entire sets of buildings. Inside the buildings you replace your bridges and routers with a switched architecture.

Here we see two examples of what can be deployed used central star wiring and switch connections to the desktop. On the left, what we are showing is a central switch within the building feeding different hubs that would provide the Local Area Network connectivity on each floor. On the second portion, we are looking at using fiber to bypass one step in the equipment. This central switch with a fiber connection allows the new technology to be moved in and upgraded without disturbing the end user connection or the infrastructure that is inside the building.

The third step is to install a fiber star backbone using current equipment. In this diagram, what we are actually showing is a FDDI ring tying together two collapsed areas within a campus. It is still being used as a backbone network; however, much of the traffic on your campus backbone is now going across the backplaying of these routers The small routers and bridges that were in the buildings have now been centralized and been replaced with switches, providing the same capability at a lower cost.

And finally, we have installing an ATM backbone. This point may be a controversial one, but we see the switching fabric as the technology of the future, providing routing at the edge, routing between the switching fabric and some legacy networks, but using switched infrastructure from the backbone down to the building. This allows the individual user to have a switched Ethernet connection at 10 Mbps, connected to a building switch providing a high-speed backplaying and then on into the ATM infrastructure, so that an end user can get a true 10 Mbps feed from his desk to any other user�s desk.

Some alternatives in collapsed technology are shown here. What we are looking at is the difference between current wiring methods, where, from an ATM switch to an Ethernet switch, fiber is used, and then within a building, copper is used down to the desktop. An even more centrally collapsed backbone would consist of the ATM fabric and Ethernet switches being centralized, and using the capabilities of multi-mode fiber to connect end users at distances of up to 2 kilometers away. The use of these differing typologies depends upon your individual campus. If your infrastructure is in place, you have wiring inside your buildings, then the first alternative will make sense. If you have a small campus with distances not exceeding 2 kilometers, providing fiber to the desktop is a long-term solution that new technologies can use for the next 10 or 15 years.

In summary, collapsing the campus backbone provides the groundwork for high-speed campus network. It avoids many of the problems with current architectures; shared band width, latency issues, and the problem with multiple network device locations. And finally, it provides a stepping stone to the new technologies which do support voice, video, and data integration.

About CREN

© CREN, 1999

Contact Us