Design of a Test-Bed System for Multimedia Uplink UCC Traffic Management in Heterogeneous 3G/4G Wireless Networks

Abstract

Focusing on enriched multimedia user created contents (UCC), also known as user-generated contents, we provide a novel uplink traffic management scheme which consists on the both of network selection and traffic redistribution. A recent trend in the use of the multimedia Internet exhibits rapid growth of UCC with the potential to create a huge amount of uplink traffic for wireless operators. Our objective is finding the optimal heterogeneous network operation framework under the precise uplink UCC traffic analysis. We suggest the market-based evaluation of the relative value of resources for each network, and develop a whole new uplink UCC operation framework in heterogeneous 3G/4G networks. To prove the applicability of the proposed operation framework and analysis, we build a large-scale uplink test-bed system. The test-bed contains all necessary functions of network facilities and protocols for development of emerging multimedia UCC traffic. We can easily test the various service scenarios in heterogeneous network environment by the developed test-bed system.

Appendix

We use the marginal cost function to determine how to transit user traffic among networks. The only one required characteristic of marginal cost function is simple strictly increasing. As the one of representative strictly increasing functions, we can consider the linear function of traffic load (traffic load = current traffic volume/total capacity) as follows:

1.1 Marginal Cost Function

$$ \begin{aligned} {\text{Marginal}}\;{\text{Cost}} = & f\left( {{\text{current}}\;{\text{traffic}}\;{\text{volume}},\;{\text{total}}\;{\text{capacity}}} \right) \\ = & a \times \left( {{{{\text{current}}\;{\text{traffic}}\;{\text{volume}}} \mathord{\left/ {\vphantom {{{\text{current}}\;{\text{traffic}}\;{\text{volume}}} {{\text{total}}\;{\text{capacity}}}}} \right. \kern-\nulldelimiterspace} {{\text{total}}\;{\text{capacity}}}}} \right) + b \\ = & a \times \left( {{\text{traffic}}\;{\text{load}}} \right) + b. \\ \end{aligned} $$

The summation of total integral function value of the marginal cost function means the total network access cost. Because the given marginal cost function is a linear function, the network access cost function is a quadratic function. The minimum value of the quadratic function, the same as the minimum value of network access cost, is obtained at the balanced point of traffic load for each cell. This example shows that the network operation policies could be simply applied by the appropriate design of marginal cost function. The generalization of mathematical derivation is as follows:

Let G = (B, C) be a connected graph with a set B = {B _i } _i  = 1,…,N = {(f _i , u _i )} _i  = 1,…,N of N cells. f _i (u _i ) is the marginal cost of traffic load u _i for the cell B _i . C = (c _ij ) matrix shows the adjacent relationship between cells.

$$ c_{ij} = \left\{ {\begin{array}{*{20}c} {1,{\text{ if}}\;{\text{cell}}\;B_{i} \;{\text{is}}\;{\text{a}}\;{\text{neighbor}}\;{\text{of}}\;{\text{cell}}\;B_{j} } \\ {0,{\text{ otherwise}} . } \\ \end{array} } \right. $$

Suppose that the maximum capacities of B ₁, B ₂, …, B _N satisfy the following relations (1) and f _i is a strictly increasing function.

$$ {{{\text{max\_capa}}\left( {B_{1} } \right)} \mathord{\left/ {\vphantom {{{\text{max\_capa}}\left( {B_{1} } \right)} {r_{1} }}} \right. \kern-\nulldelimiterspace} {r_{1} }} = {{{\text{max\_capa}}\left( {B_{2} } \right)} \mathord{\left/ {\vphantom {{{\text{max\_capa}}\left( {B_{2} } \right)} {r_{2} }}} \right. \kern-\nulldelimiterspace} {r_{2} }} = \ldots = {{{\text{max\_capa}}\left( {B_{N} } \right)} \mathord{\left/ {\vphantom {{{\text{max\_capa}}\left( {B_{N} } \right)} {r_{N} }}} \right. \kern-\nulldelimiterspace} {r_{N} }} $$

(A-1)

In view of (1), we simply set the marginal cost functions of B ₁, B ₂, …, B _N as following (2).

$$ f_{ 1} \left( u \right) = f\left( {{u \mathord{\left/ {\vphantom {u {r_{ 1} }}} \right. \kern-\nulldelimiterspace} {r_{ 1} }}} \right), \, f_{ 2} \left( u \right) = f\left( {{u \mathord{\left/ {\vphantom {u {r_{ 2} }}} \right. \kern-\nulldelimiterspace} {r_{ 2} }}} \right), \ldots , \, f_{N} \left( u \right) = f\left( {{u \mathord{\left/ {\vphantom {u {r_{N} }}} \right. \kern-\nulldelimiterspace} {r_{N} }}} \right) $$

(A-2)

Note that the marginal cost functions (2) are equivalent to the condition (3).

$f_{1} , \ldots ,f_{N} \ge 0$ & Strictly Increasing

$$ f_{1} \left( {r_{1} u} \right) = f_{2} \left( {r_{2} u} \right) = \ldots = f_{N} (r_{N} u)_{ } {\text{for}}\;{\text{all}}\;u \in [0,\infty ) $$

(A-3)

The equilibrium state ($u_{1}^{*} ,u_{2}^{*} , \ldots ,u_{N}^{*}$) should satisfy the following condition (4).

$$ f_{1} \left( {u_{1}^{*} } \right) = f_{2} \left( {u_{1}^{*} } \right) = \ldots = f_{N} \left( {u_{N}^{*} } \right) $$

(A-4)

Since f is a one-to-one function, we retrieve the following (5) which is consistent with (1).

$$ f\left( {{{u_{1}^{*} } \mathord{\left/ {\vphantom {{u_{1}^{*} } {r_{1} }}} \right. \kern-\nulldelimiterspace} {r_{1} }}} \right) = f\left( {{{u_{2}^{*} } \mathord{\left/ {\vphantom {{u_{2}^{*} } {r_{2} }}} \right. \kern-\nulldelimiterspace} {r_{2} }}} \right) = \ldots f\left( {{{u_{N}^{*} } \mathord{\left/ {\vphantom {{u_{N}^{*} } {r_{N} }}} \right. \kern-\nulldelimiterspace} {r_{N} }}} \right) \Rightarrow {{u_{1}^{*} } \mathord{\left/ {\vphantom {{u_{1}^{*} } {r_{1} }}} \right. \kern-\nulldelimiterspace} {r_{1} }} = {{u_{2}^{*} } \mathord{\left/ {\vphantom {{u_{2}^{*} } {r_{2} }}} \right. \kern-\nulldelimiterspace} {r_{2} }} = \ldots = {{u_{N}^{*} } \mathord{\left/ {\vphantom {{u_{N}^{*} } {r_{N} }}} \right. \kern-\nulldelimiterspace} {r_{N} }} $$

(A-5)

The equilibrium state (A-5) is consistent with the relations of the maximum capacity of each cell (see A-1). Also, the marginal cost functions are strictly increasing functions. Then, the minimum of total traffic load is obtained at the equilibrium state. The transition law (A-3) governs the traffic dynamics. If the marginal cost in the cell B _i is greater than the average of the marginal costs of all neighboring cells, then traffic transits from the cell B _i to the neighboring cells. After the sufficient traffic transition, we can achieve the equilibrium state (A-5) and traffic load of each cell is balanced. This effect of the transition law can be proved by the following (A-6). Let V(t) be the difference among marginal costs, V(t) is converged to zero value.

$$ V(t) = \sum\limits_{i,j} {c_{ij} \left( {f_{i} \left( {u_{i} (t)} \right) - f_{j} \left( {u_{j} (t)} \right)} \right)^{2} } \to 0 $$

(A-6)