Previous Table of Contents Next

Network Design in the Real World: Spanning Tree

Spanning tree is perhaps one of the most difficult considerations in network design. This is not due to the protocol or function per se, but rather the need for designers to consider the Layer 2 topology when incorporating Layer 3 functions, including HSRP. It is easy to create an efficient Layer 2 architecture and a separate Layer 3 design, but the two ultimately must map together to be manageable and practical. One technique is to make the HSRP primary for the VLAN root bridge. However, there are other techniques, including defining multiple default gateways on each host or using proxy ARP.

As of this writing, a new committee was meeting to design a new, faster Spanning-Tree Protocol. This protocol will likely reduce the shortcomings of the original specification, which was never designed to support today’s higher speed networks. However, as presented in the main text, the real issue is whether to design loops into the Layer 2 network at all.

At present, one school of thought on the subject is to avoid loops whenever possible and use Layer 3 routing to provide redundancy—technologies such as HSRP and MPLS (Multiprotocol Label Switching) allow fault tolerance and switching of Layer 3 packets. The other school of thought believes that spanning tree is still useful but that new features must be added to make it work in today’s networks. Cisco has a number of features that work toward this option, including PortFast and UplinkFast.

PortFast is used on switch ports that connect to a single workstation. Under this scenario, the port cannot participate in a loop, so the port should not have to go through a listening-and-learning mode. The port should also not go into blocking—there is no loop potential at this point in the network. It is important to note that this does not disable spanning tree—it simply activates the port faster than the 30-second listening/learning process would require. This feature is recommended for workstations (some of which can fail authentication to the network while the port is blocked). However, a major caveat must be added—the port cannot be connected to a hub or switch. This rule will prevent the loop creation that spanning tree was designed to prevent.

The second feature, UplinkFast, was designed to activate the blocked link quickly in the event of primary failure. Again, there are drawbacks to this feature, but when properly implemented it can greatly extend the functionality of Layer 2 loops and loop protection.

The Role of ATM

Asynchronous Transfer Mode (ATM) has been the networking technology of the 1990s. Merging the historical divisions between data, voice, and video, ATM was designed and marketed to replace all other technologies in both local and wide area networks.

At the end of the 1990s, it appeared clear that replacement of existing networks would not occur. Rather, another evolution—merging ATM with legacy technologies such as Ethernet—will likely color network design theories into the next century.

However, even with the introduction of 10Gbps Ethernet, there are still situations in which ATM can and should be deployed. Such situations include both LAN and WAN environments.

ATM operates via fixed-length cells. This design contrasts with the variable-length frames found in Ethernet and other technologies. Fixed-length cells provide consistent buffering and latency—allowing integration between voice (constant bit rate) and data (variable bit rate). ATM operates over permanent virtual circuits and switched virtual circuits.

As noted previously, ATM uses a fixed-length cell transport mechanism. These cells, at 53 bytes, are substantially smaller than the frame sizes used by Ethernet, Token Ring, and FDDI. In order to migrate between frames and cells, ATM devices perform segmentation and reassembly (SAR). The SAR function frequently became a bottleneck in older switches; however, this overhead is a minor factor today. Designers should discuss SAR processing (cells/frames per second) with their vendors before selecting a product.

ATM is often used in modern network design for WAN links and the integration of voice and data circuits. This type of installation is similar to multiplexing. In the LAN environment, ATM and ATM LANE installations are frequently used for high-speed campus backbones. This design provides a migration path for pushing ATM toward the desktop. ATM is one option for designers wishing to replace aging FDDI rings.

ATM in the LAN with LANE

LAN deployments of ATM almost always take advantage of LANE, or LAN Emulation, to integrate legacy topologies with ATM. It is unlikely that any organization would allocate sufficient funds to replace their entire existing infrastructure without some migration phase.

LANE was covered in some detail in Sybex’s Cisco LAN Switching Course Study Guide. This section will present an overview of that material for those preparing for the CID exam before the CLSC exam.

LANE makes use of at least three separate logical processes: the LAN Emulation Client (LEC), the LAN Emulation Server (LES), and the broadcast and unknown server (BUS). A fourth resource is optional but recommended. The use of the LAN Emulation Configuration Server (LECS) can greatly simplify the administrative effort needed to deploy LANE.

LAN Emulation Client

The LAN Emulation Client, or LEC, is responsible for data forwarding, address resolution, control functions, and the mapping of MAC addresses to ATM addresses.

LECs are devices that implement the LANE protocol; they may be ATM-equipped workstations, routers, or switches. It is common for an LEC to be a single element on a switch serving numerous Ethernet or Token-Ring ports. To the ATM network, it appears that the single ATM LEC is requesting data—in actuality, the LEC is simply a proxy for the individual requests from the legacy nodes.

LAN Emulation Server

The LAN Emulation Server, or LES, is unique to each ELAN (emulated LAN). The LES is responsible for managing the ELAN and providing transparency to the LECs.

Given the interdependency of the LES and BUS services, most references use the term LES/BUS pair to denote the server providing these services.

Broadcast and Unknown Server

Broadcasts and multicasts are quite common in the traditional LAN environment. Since all stations, even in Ethernet-switched installations, receive all frames destined for a MAC address containing all ones, this process works quite well and serves many upper-layer protocols, including the Address Resolution Protocol, for example.

Previous Table of Contents Next