06-03-2013, 10:54 AM
Understanding the Effects of Hotspots in Wireless Cellular Networks
Understanding the Effects.PDF (Size: 230.93 KB / Downloads: 38)
Abstract
In this work, we study and quantify the effects of
hotspots in wireless cellular networks. Hotspots are caused when
the bandwidth resources available at some location in the network
are not enough to sustain the needs of the users, which are then
blocked or dropped. A deeper understanding of hotspots can help
in conducting more realistic simulations and enable improved
network design.
We identify some causes for the formation of hotspots and
based on them, categorize hotspots into three different types:
a) capacity based, b) delay based, and c) preferential mobility
based. We show how these types have different effects on
network performance. We also consider the effects of hotspots
from various perspectives such as the number of hotspots, the
placement of hotspots, etc.
We also develop a fluid flow model and an analytical model
to study hotspots. The fluid flow model is surprisingly simple yet
effective in helping us understand hotspots and their properties.
We also describe an analytical model in which we consider a cell
as an M/M/B/B queue. We use these models to substantiate some
of the observations from the simulations.
INTRODUCTION
We study hotspots in wireless cellular networks and quantify
their effects on network performance. Hotspots can occur
whenever there is contention among users for the bandwidth
resources at some location in a network. This could potentially
lead to blocked and dropped users and thus impact the performance
of the network. Understanding and modeling hotspots
is important for conducting realistic simulations.
Hotspots have not been studied extensively in the past.
There has been some work that considers hotspots with
reference to load balancing and congestion control in wireless
cellular networks ([4], [7], [9], [10], [11]). Most of them focus
on algorithms and techniques to improve the capacity and
performance of the network in the presence of hotspots. There
has been a lack of research that specifically studies properties
of hotspots in detail.
BACKGROUND AND MOTIVATION
Background
A wireless cellular network consists of a group of cells
covering a geographical area [8]. Each cell has a base station
which is responsible for bandwidth management amongst
the users in that cell. A new user enters the network in
some random cell if there are sufficient bandwidth resources
available in that cell; otherwise, it is blocked. Once in the
network, the user keeps moving from one cell to another while
spending some time in each cell depending on its mobility
model. The time spent in a cell is referred to as cell latency
or cell residence time. During the course of its movement, if a
user is unable to move to another cell due to lack of resources,
it is dropped from the network.
Hotspots occur when there is contention for bandwidth
resources at some geographical location in a network and the
currently available resources are not enough to sustain the
demand from the users. This could potentially lead to users
being blocked or dropped from the network. We refer to such a
location in the network as a hotspot. Note that this is different
from the notion of WiFi hotspots which are locations where
wireless connectivity is available [12].
Related Work
Hotspots occur due to a difference in the load in different
parts of a network. Most researchers assume homogeneous
traffic which does not lead to an imbalance in the overall
load and no part of the network is overly loaded compared
to other parts and therefore, there is no potential for the
occurrence of hotspots. However, in real networks, traffic is
more heterogeneous than homogeneous and there is a finite
probability of hotspots. This has been recognized by the
research community and there have been some studies that deal
with hotspots in the context of load balancing or congestion
control in wireless cellular networks ([4], [7], [9], [10], [11]).
We now describe some work which, although not directly
related to the specific issue of modeling or implementing
hotspots, could be of interest to the reader. In the papers
described below, the actual implementation of a hotspot is
not very clear. The researchers simply increase the traffic load
in the hotspot cell. It is not always clear as to how this is
done, i.e. whether they increase the arrival rate of users into
the hotspot cell, decrease the departure rate of users from the
hotspot cell, increase the bandwidth demand of the existing
users in the hotspot cell, or use some other method.
Placement of hotspots
Having seen how the number of hotspots affects performance,
now we try to answer the question: Does it matter
where in the network we place a hotspot? Figure 5 shows the
result of placing a hotspot in three different regions - T1, T2,
and T3. Recall that we identified these regions in a bounceback
network in Figure 2. The hotspot in this particular case
is a delay based one.
The plots in Figure 5 show how utilization is different
depending on where one places the hotspot. A hotspot in T3
region decreases utilization the most whereas a hotspot in T1
region decreases utilization the least. This can be explained by
the fact that cells in T3 have more neighbors and hence, are
likely to have more users than cells in T1 or T2. Therefore,
choosing a hotspot in T3 region will have the most impact on
utilization since it will affect more users.
Simulations validate our model:
We ran simulations to verify this and the result is shown in
Figure 16. The plot shows that there are three distinct classes
of users and the number of users in these classes are 33,49,
and 65 which is in the ratio 2:3:4. This is the same as that
obtained in Eq.(8). This implies that the fluid flow model does
indeed capture the effects of the user movements in such
a network and can predict the distribution in the different
regions. Knowing the proportion of users in the different
regions, we can say that a hotspot in T3 region will reduce the
utilization of the network the most since T3 cells have more
users than any other type. This is the same conclusion that
we arrived at using our simulation experiments. This suggests
that the fluid flow model describes well the user distribution
in the various regions and can be used to explain the effect of
placing hotspots in different regions of the network.