How far can we go? Towards Realistic Software-Defined - HAL

How far can we go? Towards Realistic Software-Defined - HAL
How far can we go? Towards Realistic Software-Defined
Wireless Networking Experiments
Ramon Dos Reis Fontes, Mohamed Mahfoudi, Walid Dabbous, Thierry
Turletti, Christian Rothenberg
To cite this version:
Ramon Dos Reis Fontes, Mohamed Mahfoudi, Walid Dabbous, Thierry Turletti, Christian
Rothenberg. How far can we go? Towards Realistic Software-Defined Wireless Networking
Experiments. The Computer Journal (Oxford), Oxford, 2017. <hal-01480973>
HAL Id: hal-01480973
https://hal.inria.fr/hal-01480973
Submitted on 2 Mar 2017
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
1
How far can we go? Towards Realistic Software-Defined Wireless
Networking Experiments
Ramon dos Reis Fontes
Mohamed Mahfoudi, Walid Dabbous, Thierry Turletti
University of Campinas (UNICAMP), Brazil
Université Côte d’Azur, Inria, France
Christian Rothenberg
University of Campinas (UNICAMP), Brazil
Software-Defined Wireless Networking (SDWN) is an emerging approach based on decoupling radio control functions from the
radio data plane through programmatic interfaces. Despite diverse ongoing efforts to realize the vision of SDWN, many questions
remain open from multiple perspectives such as means to rapid prototype and experiment candidate software solutions applicable
to real world deployments. To this end, emulation of SDWN has the potential to boost research and development efforts by re-using
existing protocol and application stacks while mimicking the behavior of real wireless networks. In this article, we provide an
in-depth discussion on that matter focusing on the Mininet-WiFi emulator design to fill a gap in the experimental platform space.
We showcase the applicability of our emulator in an SDN wireless context by illustrating the support of a number of use cases
aiming to address the question on how far we can go in realistic SDWN experiments, including comparisons to the results obtained
in a wireless testbed. Finally, we discuss the ability to replay packet-level and radio signal traces captured in the real testbed towards
a virtual yet realistic emulation environment in support of SDWN research.
Index Terms—Mininet-WiFi; Software-Defined Wireless Networks; OpenFlow; IEEE 802.11
I. I NTRODUCTION
WiFi is ubiquitous and has become an indispensable part
of our daily lives. IEEE 802.11x standards are present in
almost any portable devices and allow rolling out wireless
infrastructures based on access point devices at a relative
low cost. While the IEEE 802.11x standards keep steadily
evolving, there are still many structural barriers to openness
that prevent the wireless infrastructure to adapt, be customized,
allow for differentiation, and leverage all wireless capacity
around us as consequence of being an infrastructure closed to
innovation [1].
It is believed that the concept of Software-Defined Networking (SDN) [2] applied to wireless networks, commonly referred to as Software-Defined Wireless Networking (SDWN) [3], [4], can break down current structural
barriers and contribute to a more innovative ecosystem. In
spirit of (wired) SDN principles, SDWN decouples control
and data planes, enabling the wireless network to become programmable by abstracting the underlying infrastructure from
applications and network services that are offered with higherlevel APIs. As a consequence, wireless network infrastructures
are presented with new deployment paths of research ideas
around seamless communication over heterogeneous wireless
networks [5], high transferring data over wireless medium [6],
energy optimization [7], among others [3], [4].
The path towards actual roll-out is not without significant
challenges, such as hardware availability, protocol implementations in software, resource constraints, realistic evaluation
models, development and testing costs and efforts, and so
on. Simulators, emulators and testbeds are common tools and
approaches used in experimentally-driven research to evaluate
the functionality and performance of networks. Aspects such
as scalability, reproducibility and cost-benefit, among others,
make the simulation and emulation the most preferred methods [8].
In the context of supporting SDWN research, a number
of alternatives have been developed such as DCE/ns-3 [9]
and OMNeT++ [10]. However, existing simulation tools face
several limitations. For instance, OMNeT++ has its own
switch/AP implementation and does not support third-party
controllers. DCE/ns-3 uses API-specific glue code for its
POSIX and kernel support, which does not cover system calls
for all applications. Another common weakness of existing
tools is the lack of 802.11 scan mechanisms which are critical
for layer-2 handover mechanisms.
Addressing the gaps in SDWN experimentation tools from
an emulation standpoint, we propose Mininet-WiFi [11] as an
emulator for SDWN that: (i) allows the execution of real code
with no modification in either kernel or applications, (ii) is
able to support best of breed open source technologies (e.g.,
OpenvSwitch, any existing OpenFlow controller), (iii) supports
the Linux mac80211 framework that allows testing most of the
IEEE 802.11 functionality leveraging user space tools (e.g.,
hostapd and wpa supplicant), among others. Mininet-WiFi is
a fork of the container-based Mininet network emulator [12]
extended to support WiFi by adding virtualized WiFi Stations
(STAs) and Access Points (APs). Mininet-WiFi inherits all
characteristics of Mininet, including the container-based emulation, a category of network emulation which allows the
creation of small emulated networks on commodity hardware
through the use of kernel level virtualization techniques [13].
Like real testbeds, Mininet-WiFi runs real code (e.g., OS
kernel, external SDN controllers applications) with real traffic
(pcap traces or actual physical, 802.11-enabled devices). Like
simulators, it supports arbitrary topologies and well-controlled
yet dynamic, real-time environments at low cost.
In this article, we pose the question on how far can we
2
go in performing realistic SDWN experiments acknowledging
that there is no tool that fits all scenarios and that the research
and deployment path is dominated by fundamental trade-offs
intrinsic to each approach (i.e., testbed, simulation, emulation).
More specifically, we try to answer the question on delivering
realistic SDWN experiments based on the design choices
and current capabilities of Mininet-WiFi, which include the
support of (user-defined) mobility models, wireless channel
propagation models, the ability to integrate with physical
devices (e.g., smartphones) [14], among other researcherfriendly features. To assess our claims of Mininet-WiFi delivering a sweet spot for realistic and reproducible SDWN
experimentation, we reproduce related work experiments, we
present a novel use case based on SSID-based forwarding, and
we carry extensive evaluation work to validate the propagation
models with traces and results from the R2lab testbed.1 We
also present the ability of replaying network conditions from
the real world by taking into account observed variations in
terms of bandwidth, latency, and packet loss.
The remainder of this article is organized as follows. The
next section provides a primer on SDWN by introducing relevant concepts to understand the applicability and relevance of
experimental tools. Section 3 discusses simulators, emulators
and testbeds as candidate platforms to support experimentation
in wireless networks. Section 4 presents the Mininet-WiFi
emulator and the main features allowing realistic SDWN
experiments. Section 5 presents the experimental outcomes of
a real testbed compared to the results obtained with MininetWiFi and the DCE/ns-3. Section 6 showcases an approach
to replay captured traces supported by Mininet-WiFi. Finally,
Section 7 concludes with some final remarks and avenues for
future work.
II. P RIMER ON S OFTWARE -D EFINED W IRELESS
N ETWORKING
Software-Defined Wireless Networking (SDWN) [3], [4] is
based on providing programmatic centralized control of the
network outside wireless-enabled devices (Access Points APs) which enforce the data plane instructions (i.e.. policy decisions) received from the controllers. The principles
of SDWN are similar to those of Software-Defined Networking (SDN) [2], i.e., a networking approach based on a programmatic separation of the control plane (aka. Network OS)
from the data plane (aka. forwarding- or data-plane). The
software-defined approach allows administrators to specify the
behavior of the network in a logically centralized manner and
at a higher-level through APIs provided by the controller platform that implements southbound interfaces to the forwarding
devices –the OpenFlow protocol [15] being the most popular
southbound interface (as illustrated in Figure 1) but not the
only one, CAPWAP [16], FORCES [17], or NETCONF [18]
are also candidate protocols in scope.
SDWN has become an emerging and significant research
branch of SDN, mainly driven by the increased interest of
mobile network operators [19], [20] and the synergies with
Network Function Virtualisation (NFV) [21]. The separation
1 http://r2lab.inria.fr
between control and data planes has existed in the wireless
domain prior to SDN and OpenFlow. Indeed, IETF standardized both LWAPP (Lightweight Access Point Protocol)
RFC5412 [22] and CAPWAP (Control And Provisioning of
Wireless Access Points) RFC4564 [16]. several years ago.
Many enterprise WLAN management systems use protocols
such as LWAPP and CAPWAP to manage their wireless
network systems. LWAPP defines the control messaging for
setup and path authentication and run-time operations whereas
CAPWAP is based on LWAPP and enables a controller to
manage a collection of wireless access points. Within the
Open Networking Foundation (ONF), the Wireless & Mobile
Working Group (WMWG) is defining a common ground architectural framework along the necessary OpenFlow protocol
extensions or enhancements to realize the identified use cases
while leveraging related work in other Standards Developing
Organizations (SDOs) such as 3GPP, IEEE, NGMN, ITU,
ETSI, IETF, etc. As per [23], over 15 use cases have been
identified, ranging from flexible and scalable packet core to
unified access networks, encompassing different elements of
OpenFlow-based or OpenFlow-oriented wireless and mobile
network domains.
SDWN research in academia has bloomed over the last
years (refer to [3] for a comprehensive survey), including
proposals such as OpenRoads [1], Odin [24], OpenRF [25],
Ethanol [26], among others. Architectures such as CloudMac [27] and Chandelle [28] use CAPWAP in their proposals.
CloudMac describes current WLAN management protocols
such as CAPWAP, as a protocol hard to extend with new
functionalities since CAPWAP AP controllers are mostly
proprietary systems. Chandelle, instead, proposes a smooth
and fast Wi-Fi roaming with SDN/OpenFlow but suffers from
Applications
(Service Providers, Operators)
Northbound interface
Wireless SDN Controllers
Control/Management of Wireless
and Mobile functions
(e.g. mobility, authorization, QoS)
Southbound interface
(e.g. OpenFlow, Capwap)
Mobile terminals
Wireless RAN
Fig. 1. Generalized and Simplified Software-Defined Wireless Networking
Architecture.
3
integration challenges with traditional switches and CAPWAP.
One issue with CAPWAP is that it tries to solve both control
and provisioning/management at once, opening the door for
conflicts due to the split of roles, e.g., consider the management layer hazards of an AP receiving a CAPWAP firmware
update command.
Identified benefits of integrating WLAN and OpenFlow [23]
are commonly related to centralized management and monitoring, unified policies, increased programmability and finegrained control of WLAN functions. Taking into account these
benefits and the limitations associated to CAPWAP –arguably
the most advanced (closed) solution today for centralizing
wireless networks management prior to SDN– some questions
are inevitable: “Is CAPWAP in scope of SDWN?”, “How to
improve the OpenFlow specification to support centralized
management of wireless networks?”, “Are radical new designs
required?” or “How much can be leveraged from currently
deployed infrastructures?”. Although these questions are still
majorly open, some noteworthy initials steps are undergoing.
There is IETF work[29] on extending the CAPWAP protocol
to support separate termination points for management, control
and data plane tunnels, and the definition of the role of an AP
and its controller(s) in RFC7494 [30]. In spirit of the longerterm mission and deliverables of the ONF WMWG, the OpenFlow protocol specification version 1.5 (section B.6.3 [31])
includes a revised behaviour when sending packets out to
incoming ports, which was a longstanding issue when mapping
wireless interfaces to switch ports.
In addition to CAPWAP-based products, there are multiple
proprietary solutions (e.g., Aerohive, Aruba, Cisco HDX, Meraki, Ruckus) based on external controllers to manage a collection of APs. These commercial solutions introduce a number
of extensions to standardized protocols or define their own
APIs between the controller and APs, presenting differences in
the refactoring of control and data plane functions in addition
to a series of proprietary radio resource enhancements. While
arguably all these solutions have proven to work well at scale,
their cost is often prohibitive for many deployments and raise
concerns due to their closely integrated nature, the consequent
vendor lock-in, and the inability for in-house or third-party
innovations.
As today, the most realistic way to experiment with WiFi
and OpenFlow together is using open source firmware and
OS solutions like OpenWRT that allows turning commodity
wireless routers into OpenFlow-enabled switches. However,
like any real testbed, such approach is subject to challenges
related to the costs and scale of the experiments, in addition
to reproducibility constraints as well as high setup times.
Wireless SDN simulators and emulators, on the other hand, are
interesting alternatives allowing to work with multiple devices
(e.g. APs and STAs) at reasonable scale in experimenterdefined environments. Before detailing our proposed research
path based on the Mininet-WiFi emulator, we next briefly
survey the main available SDWN research platforms.
III. W IRELESS S IMULATORS , E MULATORS AND T ESTBEDS
This section provides an overview of notable simulators,
emulators and testbeds identifying their main characteristics
towards a better understanding of the trade-offs when aiming
at supporting realistic SDWN experimentation.
As is well-known but commonly underestimated or misjudged when choosing an experimental platform to support
research efforts, each environment excels in some aspects
but is subject to certain limitations and/or constraints, as
depicted in Figure 3. While the exact quantification of each
characteristic and the degree of realism ultimately depend on
the accuracy of the model implemented in each specific tool
among other platform aspects that may affect each feature,
Table I (adapted from [32]) aims at illustrating the main
strengths and shortcomings typically common to each type of
experimentation approach as a first guide to choose the best
type tool for a given set of research goals and constraints.
Turning now the attention to specific wireless simulators and
emulators, Table II compares a number of features including
Type (e.g., Simulator/Emulator, Open/Close source), Programming Language (which language the solution is written) and
finally Supported Protocols (relevant protocols available by
default), Last Activity (i.e., recent updates). Similarly, Table III
attempts to categorize and compare relevant wireless testbeds
considering different criteria.
As we can see, there are a couple of alternatives for
OpenFlow-based SDWN experimentation such as DCE/ns3, OpenNet, OMNeT++, and the Estinet and Nitos testbeds.
Under the emulation/simulation space, OpenNet [41] highlights as recent work (arguably closest to Mininet-WiFi) on
combining DCE/ns-3 and the Mininet SDN emulator [12] to
provide rich SDWN experimentation features by allowing the
execution of external controllers and real applications at the
cost of a strongly coupled solution. In addition, OpenNet does
not provide high-level abstraction APIs for wireless links nor
emulation of wireless nodes (e.g., access points and stations
are not equipped with wireless network interfaces), and neither
mechanisms to select new access points before disconnection
of current link to further shorten handover latency.
As today, there is few information regarding Estinet, which
according to the developers can be used for many different
scenarios, including SDN. Being a proprietary solution and
due to its testbed nature, the availability to the wider research
community is limited. Nitos, in turn, supports four OpenFlow
switches and allows users the possibility to conduct experiments in indoor and outdoor environments.
Table IV compares Mininet-WiFi and DCE/ns-3.2 In general, in contrast to Mininet-WiFi, DCE/ns-3 does not incorporate real-world network stacks yet and might not support
execution of unmodified applications and/or without kernel
modification. Another important weakness of DCE/ns-3 is
about the development and/or improvements for IEEE 802.11.
DCE/ns-3 depends on the development of new models while
Mininet-WiFi relies on the mac80211 wireless stack of the
Linux kernel. On the other hand, DCE/ns-3 supports greater
variety of mobility and propagation models and also LongTerm Evolution (LTE) and because of this, DCE/ns-3 has been
2 Information about DCE/ns-3 relesase 1.8 were obtained from the online
manual: https://www.nsnam.org/docs/dce/manual/ns-3-dce-manual.pdf
4
ofprotocol
controller
wpa_supplicant iw iwconfig
Mininet-WiFi
nl80211
User Space
Kernel Space
station sta1
namespace
station sta2
namespace
root ap1
namespace
sta1-wlan0
sta2-wlan0
ap1-wlan0
Mobility Models
Hostapd
Propagation Models
ofdatapath
TC tool
MLME
AP mode
Configuration
Configuration management
management
for
for wireless
wireless devices
devices
wlan1
MLME
wlan3
wlan2
station mode
cfg80211
Creates Virtual WiFi Interfaces
mac80211_hwsim
mac80211
Provides MLME management services with which
drivers can be developed to support softMAC
Fig. 2. Main components of Mininet-WiFi.
TABLE I
R ANKING OF S IMULATORS , E MULATORS AND T ESTBEDS ( ADAPTED FROM [32])
Characteristic
Total Cost
Overall Fidelity
Replay Real Traces
Real Applications
Traffic Realism
Timing Realism
Scalability
Maintainability
Flexibility
Replication
Isolation
Simulators
•◦◦
•◦◦
••◦
•◦◦
•◦◦
•••
•••
•••
•••
•••
•••
Emulators
•◦◦
••◦
••◦
•••
•••
••◦
••◦
•••
•••
•••
••◦
Testbeds
•••
•••
•••
•••
•••
•••
•◦◦
•◦◦
•◦◦
•◦◦
•••
TABLE II
C OMPARISON BETWEEN M ININET-W I F I AND W IRELESS S IMULATORS & E MULATORS
Software
Type
Source
Type
Programming
Language
Mininet-WiFi [11]
Emulator
Open
Python
DCE/ns-3 [9]
Emulator
Open
C++, Python
Core[33]
Emulator
Open
several different
languages
OpenNet [34]
Simulator/
Emulator
Open
C++, Python
OMNeT++ [10]
Simulator
Open
C++
Estinet [35]
Simulator
Proprietary
?
ns-2 [36]
Simulator
Open
C++, TLC
very useful during the development of Mininet-WiFi, serving
as a basis for reference implementation.
IV. M ININET-W I F I
Mininet-WiFi is a fork of Mininet [12] extended with
the required classes to add wireless channel emulation, node
mobility, and support of 802.11 through SoftMac, a MAC layer
that provides a flexible environment for experimenting with
MAC protocols. The main components of the Mininet-WiFi
architecture are illustrated in Figure 2. In the kernel-space, the
module mac80211 hwsim is responsible for creating virtual
Wi-Fi interfaces, enabling the creation of stations and access
Supported
Protocols
Any (L3 - L7)
IEEE 802.11, 802.3,
OpenFlow
IEEE 802.11, LTE,
OpenFlow
IEEE 802.2, 802.11
IEEE 802.11, LTE,
OpenFlow
IEEE 802.11,
OpenFlow
IEEE 802.3, 802.11,
OpenFlow
IEEE 802.11, LTE
Last
Activity
2016
2016
2015
2016
2016
2015
2012
points. Still in the kernel-space, MLME (Media Access Control
Sublayer Management Entity)3 is realized by stations, while
hostapd is responsible for the counterpart task at the user-space
side in APs.
Mininet-WiFi relies on a couple of standard Linux utilities
such as iw, iwconfig, and wpa supplicant. The first two tools
are used for interface configuration and for getting information
from wireless interfaces while the latter is used with Hostapd
in order to support WPA (Wi-Fi Protected Access), among
3 Some of the functions performed by MLME are authentication, association, sending and receiving beacons, etc.
5
TABLE III
C OMPARATIVE TABLE BETWEEN M ININET-W I F I AND TESTBEDS
Software
Availability
Mobility Support
Mininet-WiFi [11]
Public
Yes
Nitos[37]
R2lab[38]
EMULAB [39]
Public
Public
Public
No
No
No
Orbit[40]
Public
No
Any (L3 - L7)
IEEE 802.11, 802.3,
OpenFlow
IEEE 802.11, WiMAX, LTE, OpenFlow
IEEE 802.11, LTE
IEEE 802.11
IEEE 802.11, WiMAX, LTE
Bluetooth, ZigBee
Mininet-WiFi
DCE/ns-3
(v1.9)
(v1.8)
3
7
sysctl, ifconfig, route
3
7
IPv6 address config.
3
7
full POSIX
3
7
poll implementation
3
3
Quagga routing stack
3
7
extensive test
7
3
real time scheduler
••◦
•••
mobility models
•◦◦
•••
propagation models
WiFi
LTE/WiFi
supported technologies
TABLE IV
C OMPARISON BETWEEN M ININET-W I F I AND DCE/ NS -3
Current mobility models supported by Mininet-WiFi
include: Random-Walk, Truncated-Levy Walk, Random4 http://tldp.org/HOWTO/Traffic-Control-HOWTO/intro.html
cost = f(CAPEX, time-to-experiment, complexity, resources, etc.)
Increased Realism/Complexity
Simulators
Less scalability, flexibility,
reproducibility, repeatibility, etc.
Emulators
Realism
Testbeds
2016
posx , posy , posz = posx + f acx , posy + f acy , posz + f acz
(1)
First, we determine the interval time (intervalT) that the
node will move to and them we calculate the motion (that includes the speed of the node) based on both initial (initPos)
and final position (finalPos) previously defined by the user
for getting what we call by motion factor (fac). Finally, we
take both initial position and motion factor in order to calculate
the next position of the node (defined and updated by pos)
until the final position.
B. Propagation Models
In the current version, Mininet-WiFi supports Free-Space,
Log-Distance, Two-Ray Ground, and International Telecommunication Union (ITU) propagation models. Other propagation models can be easily implemented by extending a single
class for propagation models. Propagation models calculate the
power of the signal received by station (RSSI) and translate
them into “equivalent” network attributes in practice, like
the maximum supported rate or the expected packet loss.
Equation 2 shows how the RSSI is calculated.
signalStrength = pT + gT + gR − P athLoss
Less real experimental
conditions
Experimental Options
2016
2016
2016
P athLoss = P ropagationM odelF ormula
Nitos
R2lab
EMULAB
Orbit
Formal Math.
Models
2016
intervalT = endT ime − startT ime
posx , posy , posz = initP osx , initP osy , initP osz
f acx = (f inalP osx − initP osx )/intervalT
f acy = (f inalP osy − initP osy )/intervalT
f acz = (f inalP osz − initP osz )/intervalT
A. Mobility
OpenNet
OMNeT++
ns-2
Last activity
Direction, Random-Waypoint, Gauss-Markov, Reference-Point
and Time-Variant Community. In addition to these readily
available model, the user is able to determine all positions
that a node have to pass through and also its speed. This is
important, because the user may have total control over the
motion of the nodes. Basically, the mobility is defined by the
Equation 1:
other tasks. Both infrastructure and ad-hoc networks are
supported. Another fundamental utility to realize the emulation
of the wireless channel is TC (Traffic Control),4 a user-space
utility program used to configure the Linux kernel packet
scheduler to control packet rate, delay, latency, and loss. TC
applies these attributes in virtual interfaces of stations and APs,
allowing Mininet-WiFi to replicate with high fidelity the actual
packet behavior as observed in the real world. The mobility and
propagation models do not require any kernel modification or
changes in the applications. More details about both mobility
and propagation models are discussed below.
Mininet-WiFi
DCE/ns-3
Core
Supported Protocols
Live
Networks
Fig. 3. Overview of related work and trade-offs of different wireless
experimental platforms.
(2)
First, the user must set the propagation model (defined by
PropagationModelFormula) to be used (free-space is
defined by default if no propagation model is set) in order
to calculate the path loss (PathLoss). Diverse parameters
are considered in the propagation models’ formula, such as
distance between transmitter and receiver, frequency, system
loss, etc. Then, the RSSI (signalStrength) is calculated
taking into account the transmission power (pT), antenna gains
of transmitter (gT) and receiver (gR) and the Path Loss.
6
The RSSI is used to calculate the maximum supported
rate during a communication between two nodes. We first
rely on the technical specification of the physical devices
(e.g., commercial wireless NIC) to define the custombw
thresholds. The link bandwidth depends on the actual signal
strength (RSSI) and the maximum supported rate for a specific
RSSI. Finally, taking into account the distance between the
transmitter and receiver, the rate is calculated through Equation 3.
rate = custombw ∗ (1.1−dist )
(3)
A similar approach is used to calculate packet loss. However, as there is no specification for actual packet loss, we
only take into account the distance between transmitter and
receiver as depicted in Equation 4.
loss = (dist ∗ 0.2)/100
WDS - Wireless Distribution System support;8 (iii) although
beacons can be captured using any packet sniffer, changes on
the RSSI by the propagation models in place are not included
in the packet header yet; and (iv) any inherited limitations from
Mininet, mainly concerning performance fidelity under high
workloads and overall scalability in terms of total amount of
nodes interfaces before performance fidelity degrades which
ultimately depends on the system platform (e.g., number of
cores, memory, etc.).
(4)
Equations 3 and 4 are based on generic approaches that
by default try to mimic the behavior presented by R2lab.
In the future, we plan to further improve the models by
supporting minstrel,5 a mac80211 rate control algorithm, and
wmediumd,6 related work that provides enhancements on the
wireless medium for mac80211 hwsim.
C. Scenario Diversity
As previously discussed, Mininet-WiFi supports both infrastructure and ad-hoc networks modes, i.e., it supports APs
where stations may connect to and stations are also able to
directly connect to each other. Both infrastructure and ad-hoc
modes can coexist in the same topology/experiment, extending the variety of possible experimentation scenarios. As an
example of ad-hoc networks, Mininet-WiFi supports VANET
(Vehicular Ad Hoc Networks) scenarios through an integration
with SUMO (Simulation of Urban MObility) providing the
node mobility patterns. Further features in scope of diverse
SDWN scenarios include support of Wi-Fi Direct (or Wi-Fi
P2P), a technology developed by the Wi-Fi Alliance to replace
the Wi-Fi ad-hoc mode, user-defined control on node mobility
to allow limiting the region of motion (even when default
mobility models are used), replaying network conditions (see
Sec. VI-C); and allowing hybrid environments blending real
physical nodes (e.g. smartphones) with the emulated environment as publicly demonstrated [14] and further discussed
in the use case experiments (see Sec. V-C). Further diverse
scenarios enabled by the experimenter-friendly features are
being driven by the user community.7
V. U SE C ASE E XPERIMENTS
This section showcases how Mininet-WiFi facilitates fast
prototyping and experimentation of SDWN use cases along
its ability to reproduce experiments from the literature. All
code and instructions to reproduce the four selected use cases
(and further ones) are publicly available in the source code
repository9 and documented in a comprehensive user manual.
A. Wireless n-Casting
This use case aims at replicating OpenFlow wireless
work [42] on bicasting over multiple wireless links. The
original experiment consisted of a video streaming application
running in a mobile station, equipped with two WiFi and
one WiMax interface, and receiving the packet flow over
multiple radios simultaneously (n-casting). The mobile station
was attached to the multiple APs using the same SSID and
OpenFlow rules were used to duplicate packets in the wired
network and re-write the L2/L3 headers at the radio access
points. As a result, the quality of experience is improved
by compensating packet loss over one wireless link by the
duplicated stream received over the alternative links(s).
In the strawman scenario shown in Figure 4(a), we use
Iperf to measure the bandwidth between STA3 and H4 during
60s. STA3 has two wireless interfaces and both interfaces
are connected to different APs (AP1 and AP2), which are
connected to an OpenFlow controller as well as the switch
S1. OpenFlow rules ensure that packets are copied and sent
through the different paths to the stations. During the first
20s the measured bandwidth is about 40Mbps, which seems
coherent with the 54Mbps limit of each interface in mode g.
When the time reaches 20s one wireless interface is disconnected (from AP2) causing a traffic decrease. The bandwidth
at the station increases again when the wireless interface is
connected back to AP2 at 40s (Figure 4(b)).
B. Multipath TCP
D. Limitations
Mininet-WiFi is under continuous development driven by
SDWN experiments that define the feature roadmap. As of
today, a number of limitations worth to note include: (i) lack
of mechanisms for channel contention (e.g. CSMA-CA), as
well as for MAC layer retransmission and interference; (ii)
We now showcase the use of MPTCP (MultiPath TCP),10 a
TCP extension that allows end-hosts to use multiple paths to
maximize network utilization and increase redundancy.
Previous work [43] presented the integration of Mininet
and physical wireless devices in order to evaluate the gains
of MPTCP in conjunction with SDN path control. We show
5 https://wireless.wiki.kernel.org/en/developers/
documentation/mac80211/ratecontrol/minstrel
6 https://github.com/bcopeland/wmediumd
7 https://groups.google.com/forum/#!forum/mininet-wifi-discuss
8 Required
improvements of mac80211 hwsim are underway;
https://www.youtube.com/watch?v=FgrESRbG61Y
9 https://github.com/intrig-unicamp/mininet-wifi/tree/master/demos
10 http://www.multipath-tcp.org/
7
AP2
(a) Topology setup
Client
mininet-wifi> sta1 ifstat
sta1-wlan0
sta1-wlan1
KB/s in KB/s out KB/s in KB/s out
246.04 5047.28 632.20 25951.69
245.37 5045.90 624.49 25949.27
246.96 5045.91 625.94 25950.82
245.09 5047.38 624.06 25949.16
245.19 5047.35 624.34 25950.53
247.08 5045.96 625.22 25949.61
245.21 5047.43 625.50 25950.95
246.84 5046.30 624.95 25948.97
244.79 5047.37 625.49 25950.64
246.59 5045.86 624.63 25950.56
(a) Sample Topology
60
50
Server
mininet-wifi> h10 ifstat
h10-eth0
KB/s in KB/s out
30592.15 886.37
30592.50 872.26
30594.53 870.65
30590.97 870.02
30594.11 870.48
30592.60 869.30
30593.90 868.02
30591.12 869.98
30594.98 871.94
30590.90 867.48
Throughput (Mbits/sec)
(b) Results
40
30
Fig. 5. MultiPath TCP
with bicasting
with bicasting
20
no bicasting
10
00
10
20
30
Time (seconds)
40
50
60
(b) Bandwidth Measurement between STA3 and H4
Fig. 4. Bicasting over WiFi
Fig. 6. Realistic experiment in hybrid physical-virtual environments
the ability of Mininet-WiFi to reproduce similar experiments
using only a virtual environment, including wireless devices,
such as laptops acting as stations and APs. Allowing APs to
be controlled by external OpenFlow controllers, different from
previous work, both wired switches and the APs are unifiedly
managed through the common flow abstraction.
The topology setup (Figure 5a) consists of 1 station (sta1)
acting as a client and 2 access points (ap2 and ap3), with
the distance between sta1 and ap2 being 8.06 meters and 1.41
meters between sta1 and ap3; 1 host (h10) acting as a server;
and 1 OpenFlow controller responsible for the flow control
based on a network-wide view. In order to obtain the results
presented in Figure 5b, we measure the bandwidth between
sta1 and h10 using ifstat to report network interfaces
bandwith for each packet transmission. As expected, the results
show that the data is transmitted by the two Wi-Fi interfaces
of sta1 (sta1-wlan0 and sta1-wlan1) and the amount of
data received by the sta1-wlan0 interface is less than sta1wlan1 (around 1.97Mbps and 5Mbps, respectively), due to
the different modes of operation (IEEE 802.11g and IEEE
802.11n) used in each interface. Following the configuration
provided by the AP TP-Link TLWR740N11 chosen for this
experiment, theoretically, up to 54Mbps over 802.11g and up
11 Mininet-WiFi
supports technical specifications of several devices
to 130Mbps over 802.11n can be provided when the signal
is up to -68 dBm. Considering the signal strength measured
at the station '-34.3 dBm over IEEE 802.11g and '-19.1
dBm over IEEE 802.11n, the maximum data rate capacity is
consumed.12 The maximum data rate is calculated by taking
into account the rate supported by TP-Link TLWR740N and
Equation 3.
C. Hybrid Physical-Virtual Environment
This use case experiment was first presented in [14] and
features a hybrid physical-virtual environment where real users
connect their 802.11-enabled smartphones to interact with the
virtualized infrastructure, including nodes forming a mesh
subnetwork, and access the global Internet after having its
traffic processed through a multi-hop OpenFlow network.
Figure 6 illustrates the scenario, where virtual and physical
mobile devices are able to interact with each other with the
switch behaviour being defined by an OpenFlow Controller.
The OpenFlow controller initially discovers the topology
and installs the required L2 flow entries to allow connectivity
12 http://www.tp-link.com.br/products/details/cat-9 TLWR740N.html#specificationsf
8
between the APs. Next, the user connects to AP1’s SSID and
tries to access any Internet Web page via HTTP. The OpenFlow controller installs one rule to re-write the IP destination
address and to re-direct all user’s HTTP traffic to a captive
portal where the user is expected to authenticate in order to
get Internet access and unlock the initial bandwidth limits
enforced through OpenFlow 1.3 metering actions. The user
can also communicate (e.g. PING) with mobile nodes forming
a mesh network.
D. SSID-based Flow Abstraction
VI. E XPERIMENTAL VALIDATION
This section delves into experiments to assess the realism of
the wireless channel emulation and overall end-to-end system
provided by Mininet-WiFi. To this end, we conduct a series
of experiments in R2lab,13 an anechoic wireless testbed that
13 http://r2lab.inria.fr/overview.md
Fig. 7. Forwarding by SSID
16
19
1
6
11
2
7
12
3
8
13
17
21
4
9
14
18
22
5
10
15
26
31
27
32
24
28
33
25
29
34
36
30
35
37
23
20
(a) Overview of R2lab’s chamber
(b) Sample Topology
Fig. 8. Experimental validation by using R2lab
will allow us to compare the results obtained in the physical
testbed to the results provided by Mininet-WiFi. In a nutshell,
−5
R2Lab Testbed
Mininet-WiFi (Free-Space)
Mininet-WiFi (Log-Distance)
Mininet-WiFi (ITU)
−10
−15
dBm
This use case illustrates the ability of fostering SDWN
research and experimentation by prototyping and trying out
ideas around SSID-based packet forwarding as illustrated in
Figure 7. This scenario is similar to the one described in
[44] but is novel in the OpenFlow implementation choice. We
mimic the case presented by the authors by assigning unique
Service Set-Identifier (SSID) for each user (or group of users)
and managing all flows through an OpenFlow controller that
defines different bandwidth limit on an SSID basis. It is very
common for an organization to have multiple SSIDs in their
wireless network for various purposes, including: (i) to provide
different security mechanisms such as WPA2-Enterprise for
your employees and an “open” network with a captive portal
for guests; (ii) to split bandwidth among different types of
service; or (iii) to reduce costs by reducing the amount of
physical access points.
Using multiple SSIDs requires the AP to map each station
to a different network connection. Traditionally, this fixed
mapping is accomplished through VLAN tagging. In our
case, we use the OpenFlow protocol to apply rules based
on input/output ports as instances of the SSIDs abstractions.
Multiple SSIDs are created in APs and each SSID is linked to
separated sub/virtual interfaces. OpenFlow rules defined how
traffic is being handled and allowed through different SSIDs.
By acting on ports, no changes were required to the OpenFlow
protocol in order to support this use case. One drawback in our
current implementation is the WiFi NIC limit of 8 sub/virtual
interfaces per virtual interface constraining each AP to handle
up to 8 different SSIDs whereas commercial WLAN solutions
are known to be able to create up to 64 SSIDs per AP.
To the best of our knowledge, this is the first time that
traffic based on a specific WLAN attribute (SSID) is defined
through OpenFlow rules. The technique is an example that
may lay the groundwork for the implementation and evaluation of more sophisticated scenarios using Mininet-WiFi, for
instance: (a) Wi-Fi Hotspots for Public Wireless Access [45];
(b) Community Wi-Fi [45]; and (c) virtual Service Providers
(vSPs) [46].
−20
−25
−30
1
2
3
4
5
log(distance) - meters
6
7
8
9 10
Fig. 9. Signal Propagation in R2lab and in different indoor propagation
models implemented in Mininet-WiFi
9
R2lab consists in a set of thirty-seven nodes on the ceiling of
a room of approximately 90m2 distributed in mesh layout to
offer an advanced simulation Wi-Fi site (see Figure 8). Being
an RF anechoic chamber built into a screened room, it provides
a suitable environment for high-fidelity, reproducible Wi-Fi
experimentation.
A. Propagation Model
In order to evaluate the propagation models implemented in
Mininet-WiFi, this case experiment uses R2lab nodes to obtain
the signal received by the stations with varying distances.
1) Approach
We first choose some nodes (3, 8, 13, 17, 21, 24 and 28)
as reference nodes, where node 3 acts as the transmitting
station and the other nodes as receivers (APs) located at
different distances.14 All the nodes were configured with
specific rate mask limited to 11Mbps, running in mode b,
channel 6 and transmission power equal to 15dBm. Then, we
used readily available instrumentation (wireshark) to record
the RSSI between the communicating nodes and compared
them with the RSSI values provided by the three supported
indoor propagation models in Mininet-WiFi. Parameters for
each propagation model include:
• Free-Space: system loss = 2
• Log-Distance: exponent = 3; system loss = 2
• ITU: power loss coefficient = 32; floor penetration loss
factor = 0; number of floors = 0
Note that variations in these parameters directly influence
the outcome of each propagation model. System loss is
a factor which is not related to propagation; exponent
represents the path loss exponent whose value is normally in
the range of 2 to 4 depending of the type of environment;
power loss coefficient represents the quantity that
expresses the loss of signal power with distance; floor
penetration loss factor is an empirical constant dependent on the number of floors the waves need to penetrate;
and number of floors represents the number of floors.
Being user-defined parameters in the propagation class of
Mininet-WiFi, the parameters can be tuned according to observed in particular scenarios with “snowflake” characteristics
(e.g., room geometry, obstacles, etc.), a fact that can be
exploited when aiming to reproduce physical experiments
using Mininet-WiFi.
2) Results
Figure 9 illustrates the results obtained for the same experiment setup in R2lab and Mininet-WiFi. We can observe
the relationship between RSSI and distance, i.e., the signal decrease as the distance between the transmitter and
receiver increases. Based on these results, we conclude
that the ITU propagation model is more appropriate. Furthermore, based on the measured RSSI values, the model
can be calibrated to reflect with more accuracy the conditions of the target environment, R2lab in our case for the
follow-up experiments. All logs from R2lab experiments can
14 Node 33 was not chosen due to observed misbehavior at the time of the
experiments.
be found at https://github.com/ramonfontes/miscellaneous/tree
/master/r2lab.
B. Simple File Transfer
We now aim at assessing the end-to-end user/application
experience in a real setup, ns-3 and in Mininet-WiFi. The
experiment consists of transferring a single file between two
nodes in the R2lab testbed (Figure 10) and measure the
transferring time, throughput, latency and packet loss. Then,
we replicate the same scenario (e.g., node distance, WiFi
modes, etc.) in Mininet-WiFi and ns-3 in order to compare
the similarity of the obtained results and hence assess the
realism that the different tools provide compared to a testbed
experiment.
1) Approach
We select node 13 as the AP and nodes 3 and 24 as the
server and client, respectively (see Figure 10). We then transfer
a 62.6MB file between server and client by using wget over
TCP and repeat this process for 10 times and measure the
total transfer time, throughput, delay, and packet loss. Relevant setup details of this experiment towards reproducibility
include: a) nodes working on channel 1, transmission power
equals to 15dBm and IEEE 802.11b enabled; b) distance
between nodes 3 and 13, and 13 and 24 equals to 2.72m and
4.08m, respectively.
2) Results
As we can observe from the results presented in Table V,
the experiment run in Mininet-WiFi yields results very close to
those obtained in the R2lab testbed. The observed difference of
the average transfer time of 10 runs (stdev) is only 5 seconds
of a total of about 180 seconds. The measured bandwidth
and end-to-end latency are also quite similar. When compared
to the results obtained in ns-3 for the same configuration15
we observe higher throughput over the wireless channel but
less goodput. Inspecting the pcap capture we observe a high
amount of TCP retransmissions and also out-of-order and
duplicated packets corresponding to approximately 16% of
the total packets transferred to complete the file transfer. The
observed retransmissions explain why ns-3 provides higher
throughput, but finalizing the file transferring in almost 4
minutes. We do not have a clear explanation for the deviations
obtained in the ns-3 results, once we tried to our best to configure the simulation parameters to match the same experiment
setup run in R2lab and ns-3.
15 Information about the file configuration can be found at the MininetWiFi’s manual
Fig. 10. Sample Topology
10
2.0
1.15
1.9
1.10
1.8
Latency (ms)
Bandwidth (Mbps)
0.45
1.20
1.05
0.40
1.00
0.95
0.35
0.90
0.85
0.30
0
20
40
60
80
Captured Traces - R2lab
Replay Captured Traces - Mininet-WiFi
0.80
100
120
140
160
180
Time (seconds)
(a) IEEE 802.11b
Captured Traces - R2lab
Replay Captured Traces - Mininet-WiFi
1.05
1.7
1.6
1.00
1.5
1.4
0.95
1.3
1.2
1.10
Latency (ms)
Captured Traces - R2lab
Replay Captured Traces - Mininet-WiFi
Bandwidth (Mbps)
0.50
0
5
10
15
20
Captured Traces - R2lab
Replay Captured Traces - Mininet-WiFi
0.90
25
30
35
40
45
Time (seconds)
(b) IEEE 802.11g
Fig. 11. Replaying Network Conditions
R2lab
Throughput
Latency
Packet Loss
Transf. Time
'337KB/s
'1ms
'0%
'3min1sec
Mininet-WiFi
(v1.8)
'352KB/s
'1.2ms
'0%
'2min56sec
ns-3 (v3.25)
'622KB/s
'1ms
'0%
3min55sec
TABLE V
M EAN VALUES OF MEASURES OBTAINED FROM TESTS
C. Replaying Network Conditions
The wireless medium is known for the frequent variations in
networking conditions change due to multiple reasons, such as
cross traffic and many sources that contribute to fluctuations of
the physical medium. We now focus on the ability of MininetWiFi to dynamically change the parameters of the network
links (e.g. bandwidth, packet loss, latency and delay) based on
the distance between the communicating nodes and eventually
augmented with observations from an actual experiment in the
real world.
Being able to replay real networking conditions based on
traffic observations in real environments is useful to predict
network performance under certain conditions, reason about
the observed network behavior, and perform fair comparisons
between alternative algorithms’ implementations subject to the
mirrors of the physical network. Previous works have explored
this approach [47], [48], some of the including wireless scenarios, e.g., TraceR [49], OMNeT++ [10], SimGrid [50], and
others. To the best of our knowledge, Mininet-WiFi pioneers
the use of this technique in SDWN.
1) Approach
In this scenario, we transfer a 62.6MB file between two
nodes in R2lab and collect real traffic using wireshark. By
filtering by TCP protocol, we record information about bandwidth and latency. We then replay these information (traces) in
Mininet-WiFi by dynamically redefining link bandwidth and
latency using Linux TC and measure the results using both
802.11b and 802.11g.
2) Results
The results are presented in Figure 11. As expected, the
total transfer time was less for 802.11g than 802.11b due
to the higher transmission speed. Mininet-WiFi outputs a
similar behaviour but with slightly higher throughput, and,
consequently, finalizes the file transfer file process before
R2lab. We believe that the difference is due to the amount
of ACK frames generated during the communication by both
receiver and transmitter, where in Mininet-WiFi those frames
are sent out to a specific interface called hwsim0 by default.
This behaviour differs from the real world where wireless
network interfaces are responsible for processing any packet
including ACK frames. If more accurate results are desired,
the parameters of the wireless channel emulation can be tuned,
for instance to reduce the bandwidth (axis X) as needed. In
the case of IEEE 802.11g, the average consumed bandwidth
in Mininet-WiFi and R2lab during the file transferring was
1.66Mbps and 1.56Mbps, respectively. Thus, reducing the
bandwidth provided by Mininet-WiFi in 0.1Mbps (i.e., about
1%) should be enough for getting further closer results when
aiming at reproducing experiments for this specific use case
scenario.
Regarding latency and packet loss both R2lab and MininetWiFi provide very similar results. Altogether, we believe that
the results are accurate within an order of magnitude, which
seems to be a “good enough” result for the wireless channel
emulation sufficient to support SDWN research commonly
focused on the higher-layer control and management features.
Nevertheless, we plan to keep improving our approach to
emulate the wireless medium with increased fidelity and
reproducibility (aka “replay”) features in order to provide more
realistic experimentation options and deliver accurate results
comparable to those from a real testbed.
VII. C ONCLUSIONS AND F UTURE W ORK
This article revolves around the question of carrying realistic
SDWN experiments, and in particular, the advances reached
by our Mininet-WiFi emulation platform.
Extending the success story of Mininet for wired SDN into
wireless, Mininet-WiFi introduces itself as a powerful tool for
SDWN research and WiFi experiments by supporting arbitrary
Linux-based applications, OpenFlow programmability, external SDN controllers, mobility and propagation models among
other experimenter-friendly features.
11
We believe that a lightweight virtualization together with
good enough wireless channel emulation capabilities, mobility
models and overall experimental scenario reproducibility are
steps forwards towards cost-effective and realistic SDWN
experimentation. Can we go further? Certainly.
With Mininet-WiFi we have achieved significant steps as
validated through the prototype use cases presented and the
proof-of-concept experiments to validate our claims of realistic
SDWN emulation. To this end we conducted experiments in
a real testbed and compared the results with those from the
Mininet-WiFi propagation models. Furthermore, we showed
the ability to replay network conditions from the real testbed to
reproduce in the emulated environment the expected behavior
from a real world experiment.
Altogether, the obtained results suggest that Mininet-WiFi
is able to mimic the real world with enough fidelity to support meaningful SDWN experiments. All necessary code and
instructions to reproduce each of the experiments presented
in this article and additional ones are available in the project
website.16
In order to reach further, some of our ongoing efforts
include (i) further validation (including scalability limits) via
experiments of more complex scenarios, (ii) improve the
replaying features of captured traces in order to provide more
accurate results and facilitate reproducibility (e.g., configure
statistics/results collection), and (iii) keep adding features and
support of scenarios as requested by the increasing user community via the mailing list and the public code repository.17
ACKNOWLEDGEMENTS
This research was partially supported by FAPESP grant
# 14/18482-4. The R2lab wireless testbed at Inria has been
funded by the ANR Equipex FIT 6165.
R EFERENCES
[1] Yap, K.-K., Sherwood, R., Kobayashi, M., Huang, T.-Y., Chan, M.,
Handigol, N., McKeown, N., and Parulkar, G. (2010) Blueprint for
introducing innovation into wireless mobile networks. Proceedings of
the Second ACM SIGCOMM Workshop on Virtualized Infrastructure
Systems and Architectures, New York, NY, USA, September VISA ’10,
pp. 25–32. ACM.
[2] Kreutz, D., Ramos, F. M. V., Verı́ssimo, P., Rothenberg, C. E., Azodolmolky, S., and Uhlig, S. (2014) Software-defined networking: A comprehensive survey. CoRR, abs/1406.0440.
[3] Jagadeesan, N. A. and Krishnamachari, B. (2014) Software-defined
networking paradigms in wireless networks: A survey. ACM Comput.
Surv., 47, 27:1–27:11.
[4] Costanzo, S., Galluccio, L., Morabito, G., and Palazzo, S. (2012)
Software Defined Wireless Networks: Unbridling SDNs.
Software Defined Networking (EWSDN), 2012 European Workshop on,
http://dx.doi.org/10.1109/ewsdn.2012.12, October, pp. 1–6. IEEE.
[5] Hossain, E., Chow, G., Leung, V. C. M., McLeod, R. D., Mišić, J., Wong,
V. W. S., and Yang, O. (2010) Vehicular telematics over heterogeneous
wireless networks: A survey. Comput. Commun., 33, 775–793.
[6] Tarokh, V., Seshadri, N., and Calderbank, A. (1998) Space-time codes
for high data rate wireless communication: performance criterion and
code construction. IEEE Transactions on Information Theory, 44, 744–
765.
[7] Khajuria, R. and Gupta, S. (2015). Energy optimization and lifetime
enhancement techniques in wireless sensor networks: A survey.
16 https://github.com/intrig-unicamp/mininet-wifi/wiki
17 https://github.com/intrig-unicamp/mininet-wifi
[8] Imran, M., Said, A., and Hasbullah, H. (2010) A survey of simulators,
emulators and testbeds for wireless sensor networks. Information
Technology (ITSim) - International Symposium, vol.2, 897–902.
[9] Mancini, E., Soni, H., Turletti, T., Dabbous, W., and Tazaki, H. (2014).
Demo abstract: Realistic Evaluation of Kernel protocols and Software
Defined Wireless Networks with DCE/ns-3. Demo Abstract in Proceedings of ACM MSWiM, Montreal, Canada, September 21-26 2014.
[10] Varga, A. and Hornig, R. (2008) An overview of the omnet++ simulation environment. Proceedings of the 1st International Conference
on Simulation Tools and Techniques for Communications, Networks
and Systems & Workshops, ICST, Brussels, Belgium, Belgium, March
Simutools ’08, pp. 60:1–60:10. ICST (Institute for Computer Sciences,
Social-Informatics and Telecommunications Engineering).
[11] Fontes, R. R., Afzal, S., Brito, S. H. B., Santos, M. A. S., and Rothenberg, C. E. (2015) Mininet-wifi: Emulating software-defined wireless
networks. Proceedings of the 2015 11th International Conference on
Network and Service Management (CNSM), Washington, DC, USA,
November CNSM ’15, pp. 384–389. IEEE Computer Society.
[12] Lantz, B., Heller, B., and McKeown, N. (2010) A network in a laptop:
Rapid prototyping for software-defined networks. Proceedings of the
9th ACM SIGCOMM Workshop on Hot Topics in Networks, New York,
NY, USA, October Hotnets-IX, pp. 19:1–19:6. ACM.
[13] Peach, S., Irwin, B., and van Heerden, R. (2016). An overview of linux
container based network emulation.
[14] Fontes, R. d. R. and Rothenberg, C. E. (2016) Mininet-wifi: A platform for hybrid physical-virtual software-defined wireless networking
research. Proceedings of the 2016 Conference on ACM SIGCOMM
2016 Conference, New York, NY, USA, August SIGCOMM ’16, pp.
607–608. ACM.
[15] McKeown, N., Anderson, T., Balakrishnan, H., Parulkar, G., Peterson,
L., Rexford, J., Shenker, S., and Turner, J. (2008) Openflow: Enabling
innovation in campus networks. SIGCOMM Comput. Commun. Rev.,
38, 69–74.
[16] Yang, L., Govindan, S., and Cheng, H. (2015). Objectives for Control
and Provisioning of Wireless Access Points (CAPWAP). RFC 4564.
[17] Doria, A., Dong, L., Wang, W., Khosravi, H. M., Salim, J. H., and
Gopal, R. (2015). Forwarding and Control Element Separation (ForCES)
Protocol Specification. RFC 5810.
[18] Enns, R., Bjorklund, M., Bierman, A., and Schnwlder, J. (2015).
Network Configuration Protocol (NETCONF). RFC 6241.
[19] Bernardos, C., La Oliva, A., Serrano, P., Banchs, A., Contreras, L. M.,
Jin, H., and Zuniga, J. C. (2014) An architecture for software defined
wireless networking. Wireless Communications, IEEE, 21, 52–61.
[20] Sama, M. R., Contreras, L. M., Kaippallimalil, J., Akiyoshi, I., Qian,
H., and Ni, H. (2015) Software-defined control of the virtualized mobile
packet core. IEEE Communications Magazine, 53, 107–115.
[21] Han, B., Gopalakrishnan, V., Ji, L., and Lee, S. (2015) Network
function virtualization: Challenges and opportunities for innovations.
Communications Magazine, IEEE, 53, 90–97.
[22] Calhoun, P. R. (2015). Lightweight Access Point Protocol. RFC 5412.
[23] Foundation, O. N. Wireless & mobile.
[24] Suresh, L., Schulz-Zander, J., Merz, R., Feldmann, A., and Vazao, T.
(2012) Towards programmable enterprise wlans with odin. Proceedings
of the First Workshop on Hot Topics in Software Defined Networks, New
York, NY, USA, August HotSDN ’12, pp. 115–120. ACM.
[25] Kumar, S., Cifuentes, D., Gollakota, S., and Katabi, D. (2013) Bringing
cross-layer mimo to today’s wireless lans. Proceedings of the ACM
SIGCOMM 2013 Conference on SIGCOMM, New York, NY, USA,
October SIGCOMM ’13, pp. 387–398. ACM.
[26] Moura, H., Bessa, G. V. C., Vieira, M. A. M., and Macedo, D. F.
(2015) Ethanol: Software defined networking for 802.11 wireless networks. IFIP/IEEE International Symposium on Integrated Network
Management, IM 2015, Ottawa, ON, Canada, 11-15 May, 2015,
http://dx.doi.org/10.1109/INM.2015.7140315, May, pp. 388–396. IEEE.
[27] Dely, P., Vestin, J., Kassler, A., Bayer, N., Einsiedler, H., and Peylo, C.
(2012) CloudMAC - An OpenFlow based architecture for 802.11 MAC
layer processing in the cloud. Globecom Workshops (GC Wkshps), 2012
IEEE, http://dx.doi.org/10.1109/glocomw.2012.6477567, December, pp.
186–191. IEEE.
[28] Monin, S., Shalimov, A., and Smeliansky, R. (2014). Chandelle: Smooth
and Fast WiFi Roaming with SDNOpenFlow.
[29] Cao, Z., Zhang, R., Hui, D., Pazhyannur, R., Gundavelli, S., Xue, L.,
and You, J. (2016) Alternate Tunnel Encapsulation for Data Frames in
CAPWAP. Internet-Draft draft-ietf-opsawg-capwap-alt-tunnel-08. Internet Engineering Task Force. Work in Progress.
[30] Shao, C., Hui, D., Bari, F., Zhang, R., Matsushima, S., and Pazhyannur,
R. (2015). IEEE 802.11 Medium Access Control (MAC) Profile for
12
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
Control and Provisioning of Wireless Access Points (CAPWAP). RFC
7494.
Foundation, O. N. (2014). Openflow switch specification - version 1.5.0.
Zimmermann, A., Günes, M., Wenig, M., Ritzerfeld, J., and Meis, U.
(2006). Architecture of the hybrid mcg-mesh testbed.
Ahrenholz, J., Danilov, C., Henderson, T. R., and Kim, J. H. (2008)
CORE: A real-time network emulator. MILCOM 2008 - 2008 IEEE
Military Communications Conference, nov. Institute of Electrical and
Electronics Engineers (IEEE).
Chan, M., Chen, C., Huang, J., Kuo, T., Yen, L., and Tseng,
C. (2014) Opennet: A simulator for software-defined wireless local area network. IEEE Wireless Communications and Networking Conference, WCNC 2014, Istanbul, Turkey, April 6-9, 2014,
http://dx.doi.org/10.1109/WCNC.2014.6953088, April, pp. 3332–3336.
IEEE.
Wang, S.-Y., Chou, C.-L., and Yang, C.-M. (2013) Estinet openflow
network simulator and emulator. IEEE Communications Magazine, 51.
Kargl, F. and Schoch, E. (2007) Simulation of manets: A qualitative
comparison between jist/swans and ns-2. Proceedings of the 1st
International Workshop on System Evaluation for Mobile Platforms,
New York, NY, USA, June MobiEval ’07, pp. 41–46. ACM.
Pechlivanidou, K., Katsalis, K., Igoumenos, I., Katsaros, D., Korakis,
T., and Tassiulas, L. (2014) NITOS testbed: A cloud based wireless
experimentation facility. 2014 26th International Teletraffic Congress
(ITC), sep. Institute of Electrical and Electronics Engineers (IEEE).
Testbeds, O. F. I. (2016). R2lab testbed. http://r2lab.inria.fr/. [Online;
accessed 27-September-2016].
White, B., Lepreau, J., Stoller, L., Ricci, R., Guruprasad, S., Newbold,
M., Hibler, M., Barb, C., and Joglekar, A. (2002) An integrated experimental environment for distributed systems and networks. SIGOPS
Oper. Syst. Rev., 36, 255–270.
Raychaudhuri, D. (2003). Orbit: Open-access research testbed for nextgeneration wireless networks.
Chan, M., Chen, C., Huang, J., Kuo, T., Yen, L., and Tseng,
C. (2014) Opennet: A simulator for software-defined wireless local area network. IEEE Wireless Communications and Networking Conference, WCNC 2014, Istanbul, Turkey, April 6-9, 2014,
http://dx.doi.org/10.1109/WCNC.2014.6953088, April, pp. 3332–3336.
IEEE.
kiong Yap, K., yuan Huang, T., Kobayashi, M., Chan, M., Sherwood, R.,
Parulkar, G., and Mckeown, N. (2009) Lossless handover with n-casting
between wifi-wimax on openroads. In ACM Mobicom (Demo.
Hyunwoo Nam, Doru Calin, and Henning Schulzrinne (2016) Towards
Dynamic MPTCP Path Control Using SDN. IEEE NetSoft, February.
Cook, C. and Schwengler, T. (2012). Simultaneous multi-mode wifi
differentiated by ssid. US Patent 8,131,303.
Bo Wang, D. A., Kenneth Wan and Thorne, D. (2015). Tr-321 - public
wi-fi access in multi-service broadband networks.
Saucez, D., Farinacci, D., Iannone, L., and Haddad, W. (2015) A virtual
Service Provider for SOHO networks. European Workshop on Software
Defined Networks, Bilbao, Spain, September, pp. 121 – 122. IEEE.
Eckhardt, D. A. and Steenkiste, P. (1999) A trace-based evaluation of
adaptive error correction for a wireless local area network. Mob. Netw.
Appl., 4, 273–287.
Noble, B. D., Satyanarayanan, M., Nguyen, G. T., and Katz, R. H.
(1997). Trace-based mobile network emulation.
Acun, B., Jain, N., Bhatele, A., Mubarak, M., Carothers, C. D., and Kal,
L. V. (2015). Preliminary evaluation of a parallel trace replay tool for
hpc network simulations.
Casanova, H., Giersch, A., Legrand, A., Quinson, M., and Suter, F.
(2014) Versatile, Scalable, and Accurate Simulation of Distributed Applications and Platforms. Journal of Parallel and Distributed Computing,
74, 2899–2917.
Rothenberg, C. E., Chua, R., Bailey, J., Winter, M., Corrła, C. N. A., de
Lucena, S. C., Salvador, M. R., and Nadeau, T. D. (2014) When open
source meets network control planes. Computer, 47, 46–54.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement