"Reduced Cell Switching in a Mobile Computing Environment"

Summary by Andy Reitz.
February 20^th, 2001

Mobile IPv4 provides a framework in which a host my change IP networks, but still maintain transparent connectivity to its home network. However, Mobile IPv4 imposes a severe penalty for mobility -- the Mobile Host (MH) must send an update message to its Home Agent (HA) every time it changes location. Since these updates are themselves IP based, they are subject to latency, loss, etc. To further compound these problems, the Mobile IPv4 standard doesn't explicitly state when a Mobile Host should change it's registration. It is conceivable that a Mobile Host may be in contact with two different access points simultaneously. In this situation, when is it beneficial to change access points (hence causing an update), and when is it prudent to stick with the current access point? This paper attempts to resolve these questions.

Cell Switching in Mobile IPv4/Mobile IPv6:

Mobile IPv4 suggests two different cell switching algorithms, but doesn't mandate a standard algorithm, nor does it mandate when such an algorithm should be used. In the first algorithm, the MH determines that it has entered a new cell if it misses several advertisements from it's old base station. In the second algorithm, the MH determines that it has changed cells by examining the IP network of the base station updates. A change in broadcast domain indicates that the MH is in a new cell, and it should attempt to establish a connection to the new base station.

One flaw with both of these algorithms is that they do not notice when the MN loses contact with the base station. Wireless communications are not necessarily symmetric -- simply because the MH is receiving updates from its chosen base station, doesn't mean that it is still able to send data to said base station. Mobile IPv6 corrects some of these deficiencies by specifying three algorithms for a base station to verify that it can reach a MH. Furthermore, two algorithms are specified for the MH to determine if it can reach the base station. Typically, of these algorithms, there are one or two well performing ones, and one slow, last resort algorithm per group. This paper seeks to improve these algorithms, to achieve less cell switches, and a stronger signal on average.

The key assumption of this paper is that in the wireless world, the coverage areas of base stations will overlap, creating an area where a MH has a choice regarding which base station it connects to. Thus, in this sort of environment, there are three simple techniques that can be used in order to prompt a change of base station:

• Late Cell Switching (Late) - Switch when moving out of the overlap area, so that contact can only be established with one base station.
• Early Cell Switching (Early) - Switch when first entering the overlap area.
• Strong Cell Switching (Strong) - By using special link-layer information, switch when the signal from the new base station becomes stronger than the current base station.

The following diagram describes these three techniques, if a MH were moving from cell A to cell B:

The performance of each of these techniques varies depending upon the movement of the MH. For example, if the MH always moves as in the picture (a user on a train), then Strong would provide the best performance. However, each of these techniques has a fairly tight degenerate region. A degenerate region is a space such that every time the MH crosses into it, a base station switch occurs. Thus, in the Early switching example, if the MH were to repeatedly cross the "westernmost extent of B" line (by moving into the overlap, back into A, and so on), a vast number of update messages would be generated. Because these degenerate regions are so easy to fall into, alternative techniques must be investigated in order to ensure adequate performance.

Drawing inspiration from the cellular telephone network (where cell switching techniques were originally pioneered), the authors observed that the technique of hysteresis could be used in this environment. Hysteresis entails inducing a switch of base station only a difference in signal strength exceeds some sort of preset value. Thus, a hysteresis region will be setup, around the area of equal strength. This region prevents excessive cell switching, as well as doesn't do so at the expense of reliable wireless communication. Using this technique, the floor is opened for three more algorithms:

• Late-SH - Late with static hysteresis. When the MH reaches a far border of the hysteresis region, it changes base stations.
• Strong-SH - Strong with static hysteresis. Maintains the "Strong" line, in the middle of the hysteresis region. Strong-SH behaves similarly to Late-SH, except that the MH will only switch cells after passing the middle line and a hysteresis boundary.
• Strong-DH - Strong with dynamic hysteresis. Same as Strong-SH, except the hysteresis boundaries change dynamically depending on MH movement.

The following diagram depicts the Late-SH technique in action:

As shown above, the MH has to cover a wide area in order to reach a degenerate state, in which it is constantly changing base stations. This is quite desirable, and will vastly improve the performance of Mobile IP.

The driving force behind this paper is that reasonable performance can be achieved with Mobile IP, if the proper cell switching techniques are used. Furthermore, the techniques specified in both Mobile IP standards are inadequate. Similarly, because wireless data networks are inherently different than the cellular voice network, the "proven" cellular techniques (Late-SH) won't provide the best performance for a Mobile IP environment, which will typically rely on smaller cells.

The authors attempt to backup these claims with some performance measurements. In these measurements, the authors find that the Late-SH technique has the minimal switching cost, it also has poor signal strength for some types of movement. Basically, with Late-SH, the MH could spend a long time close to the far edge of the hysteresis region, maintaining a poor connection to its old base station. They go on to find that the Strong techniques (Strong-DH in particular) excel at the signal quality metrics. However, they also find that Strong-DH will incur more cell switches than Late-SH. The authors conclude by stating that the extra cell switching cost is justified. Thus, the Strong-DH technique represents the best balance of signal strength versus number of cell switching costs.

1. Reduced Cell Switching in a Mobile Computing Environment
   http://dev.acm.org/pubs/articles/proceedings/comm/345910/p143-camp/p143-camp.pdf
2. Mobile Networking Through Mobile IP: A Tutorial
   http://www.computer.org/internet/v2n1/perkins.htm
3. IETF Mobile IP Working Group
   http://www.ietf.org/html.charters/mobileip-charter.html