Thanks to the complementary nature of wireless wide and local area networks, carriers are driving to merge the two within a single device with internetwork mobility. But such mobility is not without challenges. First, the wireless LAN must be able to handle session continuity, or handover, between subnets within a given network as efficiently as the wireless WAN has done for voice over the years. Second, it should be possible to reach a user device regardless of the network to which it is attached, called the roaming problem. Finally, it must be possible to maintain session continuity across the two access types.
Mobile IP has emerged over the last number of years as the solution to Internet Protocol mobility. At the Communications Design Conference, a half day tutorial (T102: "IP Mobility and the Mobile IP Technology," on Tuesday, Sept. 30) focuses on underlying issues surrounding IP mobility and how Mobile IP operates. But there's more to mobility than reaching a device and session maintenance. With emerging real time applications, it is just as important to ensure rapid handovers to avoid crippling latencies. As such, the tutorial will also describe an Internet Engineering Task Force protocol designed to enable that.
Called Fast Handover, the protocol uses IP messages alone to extend the transparent session continuity enabled by Mobile IP for a broad class of real time applications that were previously inhibited by latencies stemming from movement detection, new IP address configuration and location update. The protocol is especially suited for WLANs, in which the intrinsic handover support for IP endpoint communication is poor. As multiple versions of IEEE 802.11 technology emerge, the importance of a Layer 2 independent handover protocol such as Fast Handover becomes clear.
Fast Handover allows a wireless device to immediately regain IP connectivity after undergoing handover from a WLAN access point connected to one IP subnet to another access point linked to a different IP subnet. By design, the protocol allows a mobile node (MN) to send application packets as soon as a new link is established, and to receive packets as soon the MN's presence is detected by the new subnet router. These features effectively support such demanding time sensitive applications as real time video and voice over IP, as well as streaming media, which is less demanding but still requires continuous and relatively smooth data delivery. Furthermore, using buffering during handover improves the performance of TCP based applications, causing fewer retransmissions and reduced data latency.
The two parts of the Fast Handover solution specify how a mobile node can send packets as soon as a new link is established, and how the MN can receive packets as soon as its attachment is detected. The first part overcomes IP connectivity latency, and the second overcomes location update latency.
To establish IP communication, the MN needs to have a valid IP address. When it moves to a new point of attachment, the MN needs to know whether it has moved to a new subnet (and a new access router) as quickly as possible. Once the "movement detection" is done, the MN has to establish its new IP, or care/of, address (CoA). In the Fast Handover Protocol, the MN acquires information about its new access router prior to moving to that router. During the course of learning its surrounding access points and whether they are attached to different subnet routers, the MN builds a neighborhood data structure containing the access point identifiers and their associated access router information. This information is useful both in movement detection and in new IP address configuration.
When feasible, the MN formulates a prospective IP address for use on the target router. For instance, IPv6 stateless address autoconfiguration [RFC 2462] allows a node to configure an address without the need for a server to allocate addresses. When it is not feasible to formulate a new CoA before attaching to the new router, the protocol would still reduce the delay due to router discovery. In such a case, having the router return an address for use while the MN separately discovers local resources and other services such as DNS could reduce the address configuration delay as well.
Early acquisition of new router information reduces movement detection latency and address configuration latency. When the MN associates with a new access point, it determines whether the association implies a new subnet using the neighborhood data structure. This way, the movement detection latency can become almost insignificant. If the association with the new access point implies movement to a new subnet, the MN can start sending packets immediately using the new CoA. The stateless address autoconfiguration for IPv6 has a minuscule probability (1/264) of producing address duplicates, since there are 64 bits available for identifying the "host ID" or the interface identifier. Hence, formulating an address is straightforward, with new router prefix information provided as part of neighborhood information.
If the new access router receives a prospective new CoA as part of the Fast Handover signaling, it can check whether some other node on the new access network already claims this CoA. Alternatively, an MN can simultaneously carry out the duplicate address detection [RFC 2462] whenever DAD is required on a particular access link. For high performance during handovers, it is important to reduce or eliminate the signaling delays due to DAD, which could otherwise introduce additional latency of several seconds. With IPv4, the MN has to obtain a new CoA as soon as movement to a new subnet is detected.
To reduce the location update latency, the Fast Handover Protocol establishes a tunnel between the previous CoA and the new CoA. This tunnel is created in response to a location update message that an MN sends to its previous router. Whenever feasible, an MN sends this message prior to leaving its link with the previous router. For instance, an API from the device driver could provide a fading signal strength indication with the current access point, and identify a target access point to join. In response, the MN could decide to initiate a handover to a new access point and authorize its router to create a tunnel from the previous CoA to the new one.
If the indication is not available before communication is lost with the previous router, the MN can still send this location update message to its previous router after it detects attachment to a new router. In either case, the previous router starts tunneling packets arriving for the MN's previous CoA to its new CoA. To avoid packet loss, the new router buffers these tunneled packets until it detects that the MN is attached. The MN can expedite this detection by announcing its presence as soon as it establishes new link connectivity. To further improve the smoothness of the handover, the previous access router can also buffer packets for delayed delivery to the mobile node through the tunnel to the new CoA.
Thus, the Fast Handover Protocol effectively reduces or even eliminates IP connectivity latency and disengages location update latency from the time critical path. The net result is improved protocol support for sessions in progress to maintain their performance.
Authors' note: Opinions expressed are not necessarily those of our employer.
Rajeev Koodli (Rajeev.Koodli@nokia.com) and Charles E. Perkins (Charles.Perkins@ nokia.com) are members of the research staff at Nokia Research Center (Mountain View, Calif.).
See related chart
