China Communications • July 2017
119
edge set at time t respectively. Let M denote the
set of messages published. The system model of
data dissemination in opportunistic networks is
GHSLFWHGDV)LJ6SHFL¿FDOO\IRUDPHVVDJHm
(
), the node set are composed of publish-
ers, subscribers and relays. To describe the mod-
el, we give the following concepts.
Definition 1. &KDQQHO is a predefined
theme, denoted as C. Let ȍ denote the set of
channels, namely ȍ= {C
1
, C
2
, ..., C
S
}, where
SLVWKHQXPEHURIWKHSUHGH¿QHGFKDQQHOV
'H¿QLWLRQ Data attribution describes the
correlation between data content and the pre-
GH¿QHGFKDQQHOV7KHDWWULEXWHRIGDWDm is de-
noted by vector A
m
, A
m
= [p
1
, p
2
, ..., p
S
], where
p
i
i6p
i
LVDSUREDELOLW\YDOXHWKDW
depicts the relativity of data m and channel C
i
.
'H¿QLWLRQ1RGHLQWHUHVW describes how
much a node is interested in the predefined
channel. The interest of node v is denoted by
vector I
v
, I
v
= [p
1
, p
2
, ..., p
S
], where p
i
i6
p
i
GHVFULEHVWKHGHJUHHZKLFKQRGHv is
interested in channel C
i
.
Definition 4. Interest degree of node v
to data m, denoted as
, is the similarity of
the interest vector I
v
and attribute vector A
m
,
namely
= .
In this paper, we use the similarity formula,
i.e.
.
When
, node v is interested in data m,
where
is a threshold of interest degree.
In the system model, we assume that nodes
know the predefined channel set. Publishers
generate data but do not need to know their sub-
scribers. Thus, they only create the data attribute
for their data and inject them into networks.
Subscribers also do not know where is their
interest data and only propagate their interests
in the network. Relays make caching strategies
and forward the data close to their subscribers
according to node interest and data attribute.
A U7/7<BASED BUFFER
0$1$*(0(17P2/&<
,QWKLVVHFWLRQZH¿UVWSURSRVHWKHXWLOLW\DF-
[7,8,9]. However, more and more realistic
applications in opportunistic networks, such
as diffusing advertisements, propagating news
and disseminating traffic information, etc.,
KDYHQRVSHFL¿FGHVWLQDWLRQDGGUHVVHVLQGDWD
packets. In fact, data dissemination, which
completely decouples data producers and
FRQVXPHUVLQWLPHVSDFHDQGFRQWUROÀRZLV
more suitable for the content-centric and in-
termittent connection network [10, 11]. There-
fore, it is more meaningful to design buffer
management policies for data dissemination.
In this paper, a utility-based buffer man-
agement (UBM) policy was proposed for data
dissemination in opportunistic networks. In
8%0ZH¿UVWFRPSXWHXWLOLW\YDOXHRIFDFK-
ing data based on node`s interest and data
delivery probability, and then according to the
utility design the overall buffer management
policies including caching policy, passive and
proactive dropping policy, and scheduling pol-
LF\6SHFL¿FDOO\DWKHFDFKLQJSROLF\PHDQV
that receivers decide which messages should
be cached for improving caching efficiency;
(b) the passive dropping policy means that
which messages should be discarded when
node buffers overflow; (c) a message will be
dropped proactively to vacate buffer when its
utility are reduced to a threshold; (d) a sched-
uling policy for senders is indirectly imple-
mented by receivers via the forwarding priori-
ty of messages from senders. Simulated results
show that, compared with some existing clas-
sical buffer management policies, UBM can
obtain higher delivery ratio and lower delay
latency at the lower network cost.
The rest of this paper is organized as fol-
lows. Section
Ċ
describes the data dissem-
ination model in opportunistic networks. In
section
ċ
, we propose the utility-based buf-
fer management policy (UBM) based on the
model in Section
Ċ
. Section
Č
evaluates the
performance of UBM via simulation.
S<67(00ODEL
Opportunistic networks can be denoted by graph
G
t
= (V
t
, E
t
), where V
t
and E
t
are vertex set and
In this paper, the
authors design a
utility-based buffer
management policy
for data dissemination
in opportunistic net-
works.