Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
WiFi Direct
Marco Conti, Franca Delmastro, Giovanni Minutiello, Roberta Paris
Institute of Informatics and Telematics (IIT)
National Research Council of Italy, Pisa
Email: firstname.lastname@iit.cnr.it
Abstract—WiFi Direct introduces new opportunities to de- by supporting both peer-to-peer and AP communications,
ploy real opportunistic networks through users’ mobile devices. promising a much easier set up and management by upper-
However, its original specification does not take into account layer applications, and regular Wi-Fi speeds up to 250 Mbps.
all the parameters that can emerge from an opportunistic
network scenario, not only in terms of technical requirements However, the original specifications of WiFi Direct does not
(e.g., available resources and connectivities) but also of users take into account all the parameters that can emerge from
characteristics and profiles, which can heavily influence the an opportunistic network scenario. Here, the system has to
system’s performances and devices’ interactions. In this work deal with both devices’ and users’ requirements, considering
we investigate the feasibility of creating opportunistic networks not only technical specifications (e.g., available resources
on top of WiFi Direct framework by analyzing the protocol’s
performances in real scenarios with a variable number of and connectivities) but also personal users’ profiles that can
mobile devices. Experimental results show the times required derived from multiple information (e.g., habits, interests, user-
to form a group of variable size and the best configurations generated contents, mobility, cooperative aptitude). To this
to support opportunistic networking operations and upper layer aim, the complexity of the network management increases
applications. while offering even more efficient and personalized services.
Index Terms—WiFi Direct, P2P, opportunistic networks, mo-
bile social networks Specifically, we move the attention from the standard notion
of network of devices to the emerging concept of network of
I. I NTRODUCTION people, in which the users, their contents and needs are the
core of the system. This follows the current trends of social
Opportunistic networks represent the natural evolution of networking applications in which users needs to exchange in-
mobile ad hoc networks (MANETs), overtaking limitations formation anywhere and anytime, always requiring an Internet
and constraints of this paradigm related to the continuous connectivity. Through opportunistic communications we can
update of highly dynamic network topologies. Opportunistic further extend this paradigm and its application scenarios by
networks takes advantage of the human mobility and the exploiting the physical interaction’s opportunities among per-
consequent network dynamism by defining new opportunities sonal devices generated by the users’ mobility, not requiring
of communication generated by the occasional encounter of any pre-existent social relationships among the involved users
users and their personal devices. In this scenario, when two and not necessarily requiring an Internet connectivity. This
users and devices come into contact, they can exploit direct new paradigm is known as Mobile Social Networks [3].
communications to exchange contents, to offer services, to Currently, a few works in literature presented real imple-
share resources and also to forward messages, even towards mentations of opportunistic networks. [4] presented WiFi-Opp,
other users/devices not currently in contact [1]. In this sce- an opportunistic networking setup based on the AP mode of
nario, available Internet connection represents just another mobile users’ devices and evaluated its performances through
communication opportunity. simulations based on real human mobility traces. We applied
Initial solutions to really implement ad hoc (and subse- a similar approach in [5], [6] through real experiments but
quently opportunistic) networks relied on WiFi ad-hoc mode we experienced issues related to real users’ mobility and
that was painful to set up (designed for expert users only) intermittent connectivity events. To the best of our knowledge,
and supported data transfer speeds at most around 11 Mbps. this is the first work on real experiments on the use of
In addition, devices configured to communicate in ad-hoc WiFi Direct for opportunistic networks based on Android
mode were not able to concurrently manage an infrastructured implementation and by using a variable number of nodes. [7]
communication (e.g., AP connections) keeping the two worlds, presented preliminary experimental results considering only
ad-hoc and Internet-connectivity, completely separated. WiFi on a two-nodes configuration and by using laptops with a
Alliance [2], by introducing Wi-Fi Direct technology and the customized implementation of WiFi Direct framework. In
related software protocol, practically overtakes these limitation order to compare our experimental results with those presented
in [7], we reproduce the same configuration with our nodes as
This paper has been partially funded by Regione Toscana under the project
SmartHealthyENV (POR-CREO FESR 2007-2013) and by Registro .it under explained in Section III. In addition, we evaluate the protocol
the project CAPP-Collective Awareness Participatory Platform. performances in different configurations involving up to six
GO, it includes also the list of peers participating to the same Go Negotiation Request
Go Negotiation Confirmation
• discover, request information, and connect to other peers; Become Go
Beacon
P2P-Device-Found B
the times experienced by the two nodes. We can note that in wifiP2PManager.connect(B)
wps wps
the autonomous case each node spends about 1s to discover the
other in all the experiments. This is related to the configuration dhcp dhcp
P(X<t)
P(X<t)
0.4 0.4 0.4
(a) Cdf Discovery (b) Cdf Group Formation (c) Cdf Discovery and Group Formation
Fig. 2.
sec
A immediately sends a Group Formation Failure (GFF)
60
to C that starts a new discovery phase and successfully
40
join the group. In this case, C requires on average 12.3s 20
3% 13% 20%
newdiscovery
•
GO-Neg-Fail
No failure
Request
Double
Device
In this case it receives a Provision Discovery Failure
error
PDF
GFF
lost
(c)
(PDF) 8s after the connection request and starts a new
discovery phase in which it receives a beacon message
from A (GO). At this point it is able to join the group.
Fig. 4. Times and errors during the three-nodes experiments.
In this case C experiences 16.1s on average to join
the group (20%). It is worth noting that, in general,
when a new group starts, all the participants remove
the other discovered devices from their internal lists,
maintaining only the group participants. Therefore, it can
happens that the third node successfully completes the
Provision Discovery phase with the GO and waits to be
accepted as member of the group, but since the other two
nodes already formed a group it is forced to start a new
discovery phase. This introduces a delay of about 60s to
complete the operation (7%).
• C sends a Negotiation Request to A while it is completing
the negotiation phase with B. In this case C receives a
P2P GO negotiation failure (GO-Neg-Fail) message from
A about 99s after the connection request. This happens Fig. 5. Six nodes: incremental join.
in 40% of the experiments, requiring on average 109s to
C to join the group.
• C does not receive any message after the connection case of simultaneous requests from three nodes to create
request. wpa supplicant notifies a Device Lost event, a WiFi Direct group. Replicating the same experiments by
which invokes a new discovery phase. This represents the using four nodes, we could expect that nodes organize in two
worst case, requiring on average 130s to join the group different groups with the same probability that they discover
(7%). each other as two separate couples. Actually, we found that
the protocol is able to successfully form a single group in 90%
Finally, we also experienced a few cases in which C fails of cases with times in the range [86.5s, 275s].
twice in the join operation. The first failure due to one of
These configurations reflect a possible situation in which
the previous cases and the second one trying to connect to
three/four friends meet in a location and their devices tries
the client node even after the group has been formed. This
to establish simultaneously an opportunistic network in order
is mainly due to a delay in wpa supplicant notifications of
to exchange contents. Even though this case is possible, it is
discovered nodes. In this case C discovers B as the first node
more likely that users and mobile devices start to discover
available and tries to connect to it before being notified of the
each other not really at the same time, but incrementally, with
GO’s presence. In those cases, the join operation can require
variable delays in the joining procedure. To test the reaction
up to 2min to complete.
of the protocol to this situation we performed the following
These experiments show the application’s performances in set of experiments.
1 1
0.8 0.8
0.6 0.6
P(X<t)
P(X<t)
0.4 0.4
A/B (A,B)
0.2 C 0.2 (A,B,C)
D (A,B,C,D)
E (A,B,C,D,E)
F (A,B,C,D,E,F)
0 0
2 3 4 5 6 7 8 10 16 32 64 128 8 10 12 14 16 22 32 64 128
t - sec t - sec
(a) Cdf Single node group’s join. (b) Cdf Incremental group formation.
Fig. 6.