Performance evaluation of softer vertical handovers in multiuser heterogeneous wireless networks

Abstract

In the future fifth generation (5G) networked society, devices will integrate heterogeneous radio access technologies (RATs) to improve the network performance and the user quality of experience. In this paper, we focus on softer vertical handover (SRVH), discussing its feasibility and its performance in a multiuser scenario. Specifically, a new taxonomy for vertical handovers is proposed to resolve ambiguities in current terminology and technical issues related to SRVH implementation are discussed. Then, a simple but accurate analytical model is proposed to evaluate the performance of SRVH and results are provided with reference to best effort services in the presence of two RATs. Two case studies are considered, a mobile controlled approach with uncoordinated RATs and a network controlled approach with coordination among RATs. Results demonstrate that SRVHs are useful to allow finer granularity in resource allocation when there is coordination among RATs, although they fail to provide throughput improvements if they are selfishly performed by mobile terminals.

Appendix

Assume two RATs, with one of them having smaller coverage. We will use C (from cellular) for the RAT with larger coverage and W (from WiFi) for the other one. Assume one PoA for the RAT with larger coverage and $N_W$ PoAs for the one with smaller coverage. Assume a number of UEs, forming the set ${\mathcal {U}}$; a subset ${\mathcal {C}}$ of them are only covered by C, while subsets ${\mathcal {W}}_n$, with $n \in [1,N_W]$, are also under the coverage of PoA n of W. Obviously, ${\mathcal {U}} = {\mathcal {C}} \bigcup {\mathcal {W}}_1 \bigcup \cdots \bigcup {\mathcal {W}}_{N_W}$.

If all UEs were in ${\mathcal {C}}$ the maximum fairness was achieved by using (10). Recalling the definition provided in (15), in such case it was $T^{(k)} = 1/\eta _{C,0}$, $\forall k \in {\mathcal {U}}$.

With the availability of the other RAT, UEs in ${\mathcal {W}}_n$ ($n \in [1,N_W]$) also benefit from the resources of $W_n$. To maximize fairness, the resources of $W_n$ are distributed to UEs in ${\mathcal {W}}_n$ following (10); thus, recalling the definition provided in (17), $T^{(k)}_{W_n} = 1/\eta _{W,n}$, $\forall k \in {\mathcal {W}}_n$, $\forall n \in [1,N_W]$. Then, if we denote with $K_{C,n}$ the portion of resources of C that is allocated to the generic UE in ${\mathcal {W}}_n$, compared to the one that it would receive if it was only connected to C, and recalling the definitions provided in (15)–(17), we have

$$\begin{aligned} \left\{ \begin{array}{ll} T^{(k)} = \overline{T_C}, &{} k \in {\mathcal {C}}\\ T^{(k)} = K_{C,n} \cdot \overline{T_C} + \frac{1}{\eta _{W,n}}, &{} k \in {\mathcal {W}}_n, \forall n \in [1,N_W] \end{array} \right. \end{aligned}$$

(18)

where

$$\begin{aligned} \overline{T_{C}} \triangleq \frac{1}{\eta _{C,0}+\sum _{n=1}^{N_w} \left( K_{C,n} \eta _{C,n} \right) }. \end{aligned}$$

(19)

The maximum fairness is achieved when all UEs perceive the same throughput, i.e., when $T^{(k)}$ is the same for all UEs in ${\mathcal {U}}$. From (18) and (19) the constants $K_{C,n}$ are calculated as

$$\begin{aligned} K_{C,n} = 1 - \frac{1}{\eta _{W,n}} \cdot \frac{\eta _{C,0}+\sum _{y=1}^{N_W} \eta _{C,y}}{\eta _{W,n}+\sum _{y=1}^{N_W} \frac{\eta _{C,y}}{\eta _{W,y}}}, \forall n \in [1,N_W]. \end{aligned}$$

(20)

If any of the constants obtained from (20), for example $K_{C,z}$, is negative, it means that fairness 1 cannot be achieved neither allocating zero resources of C to UEs in ${\mathcal {W}}_z$. In such case, UEs in ${\mathcal {W}}_z$ are only connected to $W_z$ and do not perform SRVH. Furthermore, $\eta _{C,z}$ and $K_{C,z}$ should be set to 0, and the remaining $K_{C,n}$ should be recalculated.

By setting $K^{(k)}_{C} = max \left\{ 0, K_{C,n} \right\}$ for all $k \in {\mathcal {W}}_n$, varying n, and using $\eta ^*_{C,y}$ in (20) instead of $\eta _{C,y}$, as detailed in Sect. 4.4.2, we obtain (12)–(14).