Load Balancing via Relay in Next Generation Wireless Systems

Chunming Qiao* Hongyi Wut Ozan Tonguzt

A fundamental problem in current cellular systems is lim- ited capacity. Adding to this problem is unbalanced traffic among the cells. Given the explosion of the wireless traffic, es- pecially wireless data traffic for Interneweb access, and lim- ited spectrum available for licensing, congestion will occur in some cells, resulting in blocked new calls and dropped hand- offs due to the lack of available data channels (or DCH’s). Since the locations of the congested cells vary from time to time (e.g. downtowns in the Monday morning, or amusement parks in Sunday afternoon), it’s difficult to guarantee a suffi- cient amount of resources in each cell in a cost-effective way.

In this paper, we propose to integrate the cellular infras- tructure with modern wireless/mobile relaying technologies to achieve dynamic load balancing among different cells. Our basic idea is to place a number of mobile relay stations (or MRS’s) within each cell to divert traffic in one (possibly con- gested) cell to another (non-congested) cell. An example of relaying is illustrated in Figure 1 where mobile host (or MH) X in cell B (congested) communicates with the BTS in cell A (non-congested), or MH X’ through MRS’s. As shown, each MRS has two air interfaces, the C (for cellular) interface for communications with a BTS and the R (for relaying) interface for communicating with an MH or another MRS. While the C interface may operate at 850 MHz (or 1900 MHz in PCS sys- tems or 2 GHz in 3G systems), the R interface can use an un- licensed band at 2.4 GHz (with interference cancellation tech- niques) as in wireless LANs or mobile Ad-hoc networks.

Figure 1: A relaying example.

Using MRS’s to increase the capacity of an existing cellular system is more cost-effective than installing additional BTS’s (as in cell splitting) because an MRS is a wireless commu- nication device which can be deployed more easily. In addi- tion, since multiple MRS’s can be used for relaying, the trans- mission range of each MRS using its R interface can be much shorter than that of a BTS, which means that an MRS can be much more smaller and less costly than a BTS.

*Dept. of Computer Science and Engineering, and Dept. of Electrical En-

t Computer Science and Engineering, Email: hongyiwu@cse.buffalo.edu IElecuical Engineering. Email: tonguz@eng.buffalo.edu.

gineering, University at Buffalo (SUNY). Email: qiao@computer.org.

At the same time, using MRS’s for relaying makes more practical sense than using MH’s for relaying because MRS’s (approximate) locations and their (potential) movement can be controlled, e.g., by Mobile Switch Centers (or MSC). More specifically, MRS’s can communicate with each other and with BTS’s at a higher data rate than MH’s, due to MRS’s limited mobility and possible use of specialized hardware (and power source). In addition, using MRS’s for relaying facilitates MSC to perform (or at least assist in performing) critical call man- agement functions such as authentication and billing. In par- ticular, it enables a relaying route with satisfactory QoS pa- rameters (if it exists) to be established quickly, and once es- tablished, maintained with a high degree of stability.

Among the MRS’s involved in relaying, we may call an MRS which directly communicates with an MH (e.g. MRS 1 in Figure 1) a proxy, and an MRS which directly communicates with a BTS (e.g. MRS 2 in Figure 1) a gateway (an MRS can serve as both a proxy and a gateway at the same time as illus- trated in Figure 2 (a)). When and only when an MRS serves as a gateway, it uses the C and R interfaces concurrently. Other MRS’s along a relaying route use the R interface only.

In an existing cellular system, if MH X initiates a call from a congested cell B, the call will be blocked. In the proposed next generation wireless system with integrated cellular and relaying technologies, however, MH X can switch over to the R interface to communicate with a BTS in a non-congested cell through MRS’s as in Figure 1. We call this strategy primary relaying.

Figure 2: To free up a DCH for use by MH X, one can use (a) secondary relaying from MH Y to BTS A, or (b) cascaded relaying (i.e. MH Y to BTS C and MH Z to BTS D).

If primary relaying fails, one may resort to secondary relay- ing so as to free up a DCH currently used by one of the active MH’s in cell B for use by MH X. More specifically, as shown in Figure 2 (a), one may establish a relaying route between MH Y and BTS A (or any other cell). In this way, after MH Y switches over, the DCH used by MH Y can now be used by

149 0-7803-6534-8/00/$10.00 0 2000 IEEE

MH X. Even if secondary relaying fails because, for example, cell

C that MH Y has been relayed to is also congested (as shown in Figure 2 (b)), the new call may still be supported using cas- caded relay. More specifically, one may apply secondary re- laying to free up a DCH for used by MRS G. In other words, one can establish a relaying route between an MH (say MH Z ) in cell C, and a non-congested cell (say D), or its callerkallee MH Z ’ . In this way, MRS G can be allocated the DCH previ- ously used by MH Z in cell C, and in turn MH X can be allo- cated the DCH previously used by MH Y in cell B.

Let R and r be the transmission radius of a BTS and an ’ MRS, respectively, and define n = [R2/r2] . In order to pro-

vide overlapped coverage of a cell with MRS’s so that data can be relayed between any two locations in a cell (without involv- ing a BTS), one need approximately 2n MRS’s, whose repre- sentative values are shown in Table 1.

Obviously, when M is large enough (e.g. , when M = loo), P2 approaches (1 - p ) , which implies that the probability of setting up a relaying route via either primary or secondary re- laying will approach 1, or in other words, one can almost al- ways relay a new call from cell A to Bi.

Of course, after relaying some number of new calls to B,, Bi may become congested, at which time, cascaded relaying becomes useful. This is verified by our simulations, which also show the increase in the amount of the effective capac- ity of cell A due to relaying. A location-dependent traffic pat- tern is simulated, where cell A is a hot-spot (i.e., its capacity is nearly reached), and the cells closer to the hot-spot (i.e., Bi’s) experience a heavier traffic load than those further away from the hot-spot (i.e., cells Cj’s, where 1 5 j 5 12).

The following figure shows the additional numbers of new calls that can be supported in cell A concurrently, without re- laying, and with primary only, secondary, and cascaded relay- ing, respectively.

Table 1: The Maximum number of MRS’s needed per cell

Given a limited number of MRS’s, one may place a seed MRS at every border of two adjacent cells so that it can com- municate with two BTS’s on the C-interface (see Figure 3 for an illustration). Each of the additional MRS’s will grow from the seeds in such a way that it forms a cluster involving at least one seed.

..-.

.-._

@ Coverage and position 0 Coverage and position of additional of seed MRS MRS’s grown from the seed MRS’

Figure 3: A subsystem with a limited number of MRS’s

Let p be the probability that a MH in a cell is covered by an MRS (p = 3r2/R2 if there are only seed MRS’s, and is approximately 19% when R = 2km and r = 500m). As- suming that cell A is congested but cells B,, (1 < i 5 s), are not. Then, the probability of finding a primary relaying route from an MH in cell A to one of its neighbor Bi is PI = p . In addition, let the number of DCH’s in a cell be M . Then, the (conditional) probability of finding at least one secondary re- laying route (assuming that a primary relaying route cannot be found) is given by

P2 = (1 - P l ) . [ l - (1 - p ) M ] = (1 - p ) . [ l - (1 - p ) M ]

Figure 4: Supported new calls in cell A (in addition to existing calls) as a function of the total number of MRS’s.

The simulation results are obtained by assuming that the number of DCH’s in a cell is M = 100, R = 2km and r = 500m. In addition, the average number of additinal MRS’s in each of the cells A and Bi is assumed to vary from 0 to 19 (and hence, the total number of MRS’s in the 19 cells, including the seed MRS’s in A, Bi’s and Cj’s, varies from 24 to 157). For simplicity, it is also assumed that an MRS may relay as many calls as needed although a gateway M R S relaying m calls will be assigned m channels for use on the C-interface.

In summary, the proposed integration of the cellular and mobile/wireless relaying technologies can dynamically bal- ance traffic between cells in a cellular system, increase the ef- fective capacity of the system, and circumvent congestion (in that it is possible to maintain calls involving MH’s moving into a congested cell, or to accept new call requests involv- ing MH’s in a congested cell). It can also provide unmatched routing flexibility, enhance the coverage, reliability (or fault- tolerance) and scalability of the system, and potentially im- prove MHs’ battery life and transmission rate. This novel con- cept of integrating cellular and mobile relaying technologies may lead to the convergence between the third generation (3G) wireless systems and Ad-hoc networking such as Bluetooth. It may also be applicable when building a wide range of com- munications systems including wireless local loop (WLL) and hierarchical satellite-terrestrial networks.

150