9.1 Modern Networks Use Layered Protocols


Over the past few centuries, the transportation of physical goods has evolved from using specialized railcars, wagons, and ships that were loaded manually to using universal cargo containers that can be moved over railroads, highways, and oceans without repacking. Likewise, over the past few decades, the transportation of electronic information has evolved from proprietary systems to systems that use standard protocol layers so data now can travel around the globe over various types of networks in a common IP packet.

Layer one , the physical layer , can be likened to a railroad . As long as the wheels fit the rail width, it doesn't really matter what type of railcars are used. It could be boxcars, hopper cars , or flatcars ”the railroad moves them all. Twisted-pair and fiber- optic cables, including their carrier signals, are examples of physical-layer networks, as are wave division multiplexers (muxes) . Since they also provide rudimentary framing, Synchronous Optical Network (SONET) and Synchronous Digital Hierarchy (SDH) do not fit the exact definition of a physical-layer network, but often get listed as such anyway. Perhaps they could be likened to railcars, which generally are considered to be part of the rail service. Here, too, they will be called physical-layer protocols. SONET and SDH offer only fixed-bandwidth service and connection-oriented addressing, whereas most layer two protocols offer variable bandwidth service and the more flexible connectionless addressing, so the characteristics of SONET and SDH are a bit more like layer one than layer two.

Layer two , the link layer , is analogous to big-rig truck trailers . They can be pulled on most streets and highways, moved on flatbed railcars, and even can be driven onto a ferryboat for short trips across water without unloading the contents. Ethernet, Token Ring, and Fibre Channel are examples of link-layer network protocols and can be transmitted over almost any type of physical layer network.

Layer three , the network layer , would correspond in this example to universal cargo containers. There are only subtle differences between shipping systems based on truck trailers and those based on modular cargo containers. For long-haul shipments by rail or sea, it is more efficient not to ship the wheeled trailer, so it is common to ship only the containers and place them on another wheeled trailer at the destination for local distribution. Compared to trucking systems, perhaps today's global containerized cargo-routing systems are a bit more sophisticated and universal.

That may be a good representation of the subtle differences between layer-two and layer-three networking. For long-haul data transmission, it's a bit more efficient to strip off the Ethernet frames and send only the network-layer packets, reframing them with Ethernet again at the destination for local distribution. Likewise, the layer-three network-routing algorithms are more sophisticated and universal. IP, DECnet, and AppleTalk are examples of network-layer protocols. Over the past few years , IP has become the de facto standard.

These analogies are shown in Figure 9-1. Just as cargo containers can be placed on big-rig truck trailers and driven onto flatbed railcars for long-haul transportation, IP packets can be encapsulated in Ethernet frames and transported over SONET multiplexer networks (SDH outside North America). Most SONET (and SDH) muxes now have Ethernet interface cards, so Ethernet LAN switches can be connected directly to the muxes. Extending the railroad comparison, for somewhat better efficiency, cargo containers can be placed directly on the railcars, just as IP packets can be sent directly over SONET multiplexer networks using packet-over-SONET routers.

Figure 9-1. Cargo and data transport analogies.

graphics/09fig01.jpg

Packet-over-SONET routers strip off the Ethernet frames as they are received from the LAN switches and forward only the IP packets to the multiplexers, using the SONET format. The process is reversed at the destination so that Ethernet LAN switches can be used for local distribution of the IP packets. In this case, using the designation that the SONET muxes are layer-one devices, the long-haul portion of the network does not use layer two, as the layer-three packets are sent directly over the layer-one network.

Figure 9-2 shows a comparison of the protocols that are used to transport the data end to end, along with the transmission rates commonly used today. Note that the networking industry generally expresses rates in bits per second (bps), rather than in bytes per second (Bps), as is common in the storage and computing industries.

Figure 9-2. Ethernet interfaces on multiplexers enable direct link from LAN switches.

graphics/09fig02.jpg

Now that SONET muxes have Ethernet interface cards, routers no longer are needed at each site to convert Ethernet into the SONET format. That has given rise to a new type of metro-area service, from a new category of service providers, sometimes called Ethernet Local Exchange Carriers (ELECs). The ELEC's subscribers (customers) see only Ethernet ”most likely Fast Ethernet (100 Mbps) or Gigabit Ethernet (1000 Mbps) ”rather than SONET or SDH, at each end of their connection.

For reference, here is some additional service provider lingo: The equipment provided by the subscriber is called customer premises equipment ( CPE ) . The interface between the subscriber and the service provider is the demarcation point . In this case, the ELEC's demarcation point is Fast Ethernet or Gigabit Ethernet, thereby reducing the CPE cost, as no router is needed to convert from Ethernet to SONET.

It also is possible to interconnect the Ethernet switches directly with dedicated fiber-optic cable ” especially within an office complex or a campus, where it may be somewhat easier to trench new cable or route it through underground steam tunnels. Metropolitan- and wide-area dedicated fiber is more difficult to get, since it must cross public and perhaps private property. Still, some service providers may lease unused (dark) fiber to end users or trench new fiber under city streets for a large one-time charge or a long- term lease agreement.

Even where it's available, the cost of dedicated fiber in the metropolitan and wide areas generally has remained quite high, and the installation times are excessive. Consequently, most organizations use public networks with customer interfaces provided at layer one (typically SONET or SDH), layer two (Fast Ethernet and Gigabit Ethernet, now available in many metropolitan areas), or layer three (IP). The Internet is the largest public IP network, and its performance and reliability have become quite good. In fact, virtually all organizations now use it for transmitting email messages, and small ones typically use the Internet for all applications. For critical data transmissions, however, most midsize and large organizations also build their own private IP router network, overlaid on a public SONET/SDH backbone network.

There are additional upper layers in the networking stack, such as TCP (Transmission Control Protocol) at layer four, and other actions such as load balancing. The common models extend through layer seven. However, since most MANs and WANs are generally limited to layers one through three, and layer four typically applies to end systems, the discussion in this chapter is focused on the lower layers.



IP Storage Networking Straight to the Core
IP Storage Networking: Straight to the Core
ISBN: 0321159608
EAN: 2147483647
Year: 2003
Pages: 108

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net