VLAN History The first products to feature VLANs were segmentable switches. The segmentable switch lets you create two or more virtual switches from a single switch. By assigning individual ports to a specific virtual switch, you create two or more logical switches within a single device. These first VLAN products let you switch multiple segments on a single switch--a definite advantage when early switched Ethernet ports were priced at $700 and higher.

Per-port Ethernet prices have been steadily declining as vendors integrate more of the switching functionality in the latest ASIC designs. Today, many sites have a healthy mix of high-speed switched and shared LANs. VLANs let you design multiple segments for security and broadcast containment. To extend these capabilities beyond a single switch, vendors have created proprietary protocols and signaling mechanisms that let VLANs exist among a number of swi tches. These mechanisms let customers design a network based on logical layout of an organization, rather than by physical location. Unfortunately, these solutions are proprietary.

Determining VLAN membership used to be done solely on a per-port basis. Each port could support only one VLAN. Today, VLAN membership can be determined in several ways. Many of these methods were designed to facilitate moves, adds and changes, or to provide security on the network (see "Types of VLANs" on page 178).

Whether you choose simple port-based VLANs or an elaborate policy-based scheme, the goal is the same: VLANs make it possible for your network layer topology to conform to your organizational structure rather than the physical layout of your network, in theory making it easier to administer. The features you get will depend on how much you're willing to pay and administer.

Making It a Standard The missions of the 802.1p and 802.1Q standards bodies are to provide a uniform method for conveying frame prio rity and VLAN trunking information across the network. And most of the legwork for the two standards has been completed. Two bytes have been added to the Ethernet frame. These 16 bits are dedicated to VLAN membership and priority information. Specifically, 3 bits allow for up to eight priority levels, 12 bits allow for up to 4,096 different VLANs, and 1 bit is reserved for foreign VLAN frames types--like Token-Ring and FDDI--switched over an Ethernet backbone.

But adding 2 bytes to the maximum size of the Ethernet frame introduces problems for many switches, which forward Ethernet frames in hardware. To overcome this, the standards body has proposed reducing the maximum data payload per frame by 2 bytes. Though unpopular, this may be the solution that permits legacy connectivity to older switches and routers.

Getting Your Priorities Straight An integral part of the 802.1Q development process, IEEE 802.1p is a method for signaling network priority on a per-frame b asis. Although most LANs don't see continual congestion, bursts of traffic may introduce latency. These are unacceptable in real-time networks designed to support voice and video. 802.1p specifies a method for reordering packets in a queue with frames in place, thus allowing for timely delivery of highly delay-sensitive traffic.

The Integrated Services Over Specific Link Layers (ISSLL) workgroup has defined a number of recommended service classes based on how much latency a packet can withstand. To get a general feel for these priorities, imagine different priorities for traffic with sub-10-ms delay sensitivity, 100-ms sensitivity, best effort and so on. By mapping application layer protocols, like the Resource Reservation Protocol (RSVP), on top of these service classes, a complete policy for controlling priority can be defined. Several Gigabit Ethernet newcomers, including Foundry Networks and Prominet, are promising RSVP support in current products once the standard has been completed.