Full Mobility and the OSI Protocol Stack

Full mobility can be implemented several places in the Open System Interconnection (OSI) protocol stack. Figure 1-2 shows three common divisions of the OSI stack. At the bottom is the access technology, where many existing mobility protocols reside. The layers at the top are usually controlled by software applications, where no standard mobility protocol exists. In the middle, the network or IP layer is rapidly proving itself to be the ideal place for link- and application-independent services like mobility.

Figure 1-2. Three Divisions of the OSI Stack

Mobility at Layer 2

Mobility integrated with the access technology has proven to be successful. The Layer 2 protocol is aware of all relevant changes in access links; thus, the information required for move detection and location discovery is readily available. User experience is generally positive, because handover can be quick and no impact on high-level applications exists. However, like many other aspects of Layer 2 protocols, scalability can be a problem.

Link layer mobility protocols alone are not capable of interaccess technology handover. Interaccess technology handover is especially important for data communications because of the wide range of speed and density of coverage provided by existing protocols. When a user is sitting at his desk, he wants to be able to take advantage of the high speed and low latency available through fixed Ethernet, but when he leaves his desk, he wants to maintain his connection through a WLAN. Mobility solutions at higher layers of the OSI stack are capable of providing roaming among diverse access technologies.

Mobility at Layers 47

Though there has been little effort to standardize a mobility protocol in Layers 4 through 7, many applications implement some level of mobility. Instant-messaging clients check regularly to ensure that the access link is still available and that the network address is the same. Many e-mail packages can handle IP address changes that occur between mail checks. When it comes to full mobility, most applications cannot maintain sessions when the access links change. If a user removes her fixed Ethernet cable while she is attempting to download a large e-mail attachment, the download fails. Few applications are even intelligent enough to resume or restart the transfer without error. Mobility at Layers 4 through 7 is not practical because applications must specifically be designed with mobility support. Each application must deal with mobility differently, common hooks are needed in the operating system to ensure that applications have all the information they need, and legacy applications must be retrofitted.

Mobility at Layer 3

Layer 3 (the network layer) and especially IP are ideal candidates for supporting full mobility protocols. IP is supported on a wide range of wired and wireless links, and is supported by many applications. IP mobility allows all IP-enabled applications, whether they use Transmission Control Protocol (TCP), User Datagram Protocol (UDP), or any other transport protocol, to seamlessly inherit full mobility across a diverse range of access link types. Users can roam from fixed Ethernet to wireless Ethernet to cellular, only noticing the degradation in speed and latency. The user does not need to restart applications, interrupt sessions, or reboot.

Combining Mobility Protocols

Both Layer 2 and 3 mobility solutions have their own unique advantages that dovetail nicely to create a full mobility solution. Many real-world mobility solutions rely on a combination of mobility solutions. Layer 2 mobility provides fast handover among access links in a small area and of the same technology. Layer 3 mobility is added on top of Layer 2 mobility to provide scalability and link-layer independence. Many of the current data offerings in mobile telephone systems use a combination of fast Layer 2 mobility and Mobile IP or a similar protocol for scalable Layer 3 mobility.