Active Networking based Service Infrastructure for
Cellular WLANs
Reiner N. Schmid
1
, Cornel Klein
1
, and Thomas Becker
2
1
Siemens AG, Munich, Germany
2
Fraunhofer Institute FOKUS, Berlin, Germany
Abstract. An approach of managing a service infrastructure for wireless local
area networks (WLANs) is presented. This approach uses Active Network tech-
nology as developed in the IST project FAIN in order to deploy and run services
at appropriate locations in the network. By introducing a service which monitors
traffic and adaptively controls the access of mobile users it is possible to dynam-
ically bind particular users to specific access points, thus transforming conven-
tional WLAN to cellular WLAN. This service is made up from components in
the data and control plane of the network. Additional components in the manage-
ment plane are used for the deployment and runtime management of the service
components. The usefulness of this approach is corroborated with the outline of
key scenarios.
1 Introduction
WLAN [2] networks have become very popular for a wide range of users and services
and are rapidly deployed as public hot spots in areas such as airports, conference rooms,
restaurants, shopping malls and public fairs. WLAN has emerged as the technology of
choice used for wireless data transfer in public areas. From a technological point of view
it is comparable to Ethernet as it establishes a shared medium for data exchange without
provisions for exclusive access. This concept is quite opposite to mobile communication
technologies such as GSM or UMTS, which have mechanisms for exclusive channels
and cellular structures, being the basis for the provisioning of Quality of Service (QoS)
guarantees and other mechanisms for resource control.
WLAN, as standardised by IEEE, offers only a limited support for advanced cellu-
lar mechanisms and roaming. Though there is emerging support for inter access point
communication in the recently established standard IEEE 802.11.f – defining commu-
nication between the access points about re-association of terminals this is limited
to facilitate the deployment of large scale networks consisting of many WLAN access
points.
WLAN, based on IEEE 802.11b, allows for transmission rates up to 11 Mbps, which
is sufficient for audio transmission and to a limited extent also for video streaming given
that the streaming bandwidth is adapted to the available transmission bandwidth (see
e.g. [3]). Though the newer and faster WLAN standards (e.g. IEEE 802.11g) allowing
N. Schmid R., Klein C. and Becker T. (2004).
Active Networking based Service Infrastructure for Cellular WLANs.
In Proceedings of the 3rd International Workshop on Wireless Information Systems, pages 134-141
DOI: 10.5220/0002674801340141
Copyright
c
SciTePress
for higher data rates are now available, support for dedicated Quality of Service is lim-
ited and not sufficient to support service environments with different services using the
same WLAN network at the same time.
In this paper we focus on WLAN application scenarios in semi-public areas (air-
ports, hotels etc.), which require a flexible service provisioning infrastructure concept
in order to fulfil the requirements of users and services alike. With respect to these
scenarios as discussed in section 2, more flexible and sophisticated mechanisms for
support of service infrastructure by WLAN are required. Though this isn’t part of the
MAC layer specification of WLAN by IEEE, it is shown in this paper that this ser-
vice infrastructure can be implemented by using Active Networking [1], a promising
method of solving the demands of next generation networking infrastructures. The ba-
sic concept of Active Networks is to allow third parties (service providers, end-users,
etc.) to inject application specific service code into the network. Active Network nodes
expose open interfaces and execution environments for running services, thereby mak-
ing the network ‘programmable’. Wireless local area networks (WLANs) prove to be a
particular interesting application domain for Active Networking [4].
Active Networking concepts, in particular the Active Node architecture of FAIN as
presented in section 3, are used to implement a flexible, application adaptable resource
management of WLAN networks. This includes the establishment of virtual service
environments and dynamic access control for WLAN resources thus providing cellular
mechanisms for WLAN management.
The main idea of our approach is to transfer concepts of cellular mobile networks
to WLAN establishing a cellular WLAN, thus complementing out-of-the-box WLAN
with state-of-the-art concepts of mobile networks. This is shown in Section 4 which
presents the details of a service for dynamic control of access to a WLAN. Conclusions
are discussed in Section 5.
2 Mobile Active Networking Scenarios
We mainly consider WLANs in so-called semi-public areas. By a semi-public area we
denote a geographically separated area with a network infrastructure which is populated
by a huge variety of stakeholders, but which is owned and controlled by a particular
institution, e.g. a company. In contrast to completely private areas with very restricted
access, it is frequented by a large number of users with different objectives and interests.
Some examples for such semi-public areas and the users of their WLAN infrastructure
are:
airports with users like passengers, airlines, shops, airport personnel;
trade shows with users like visitors, exhibitors, organiser’s personnel;
shopping centres providing WLAN support for visitors, shops clerks, service per-
sonnel, etc.;
hotels with guests, conference organiser, hotel personnel;
corporate campuses with employees, external visitors, different organisational units.
In all these cases there are a couple of groups with different requirements when us-
ing the WLAN infrastructure. For instance, at a trade show the visitors may prefer best-
effort or QoS-based Internet access, depending on their willingness to pay. Exhibitors
135
may want to use WLAN for providing location-based services to visitors, e.g. to attract
people to visit their exhibition places. Additionally, exhibitors could use WLAN for
their internal communication as do the employees of the trade show organiser. In that
case it is likely that WLAN based telephony will only be used if it is ensured that it
doesn’t interfere with other WLAN usage, e.g. ongoing data transfer.
To summarise, the requirements of the different users towards WLAN operation are
quite specific and diverse. Ad-hoc approaches, where different WLANs are installed
and operated by (or for) the different stakeholders aren’t suited to accommodate their
joint requirements for the following reasons:
Setting up separate WLANs is likely to be more expensive than using a shared
infrastructure, in particular, if they are used only for a short period of time.
Overall Quality of Services requirements can only be implemented if access to the
underlying shared resources as WLAN channels and bandwidth is controlled by a
common entity.
Moreover, a powerful shared infrastructure is beneficial for high-volume peak times,
as, for instance, an application with high bandwidth requirements can be shifted to
a specific WLAN channel or access point.
Since the involved stakeholders and their requirements in the presented examples
are changing quite frequently, static approaches to the configuration of such a common,
shared infrastructure aren’t well suited. In particular, it is rather unlikely that all required
services are configured by the infrastructure’s operator in advance. What we need is an
infrastructure which among others implements the following requirements:
Delegation: part of the management of the infrastructure can be delegated to its
stakeholders, allowing them to manage their part of the network on their own. The
provided management functionality should be as flexible as possible w.r.t. to the
kind of applications they have in mind.
Virtualisation: the different stakeholders shouldn’t interfere with each other, but
should get their own (virtual) private network, both w.r.t. to management function-
ality and user traffic.
Resource Management: it should be possible to flexibly allocate, release and mon-
itor the underlying shared resources on a per stakeholder/per users basis.
The Active Networks based service infrastructure presented in this paper realises these
requirements complementing existing approaches for deploying and operating large
WLANs. For this reason we don’t deal with issues like network planning (e.g. access
point placement), but we rather take for granted viable approaches for dealing with
these problems.
3 FAIN Active Node
The management layer of the FAIN Active Node [4] is of particular importance in the
FAIN architecture. The purpose of this layer is to manage active services in the scope of
a network node on the data, control, and management planes. It serves as an abstraction
layer between services and the node’s operating system by abstracting from operating
136
system specific interfaces and making them available as so called basic node services
to higher-level services. In this way the management layer provides a uniform way to
deal with functionality provided by the node as well as dynamically added extensions
in the form of active services.
Since services may be implemented using different technology and thus require
different execution environments (EEs) the FAIN node architecture doesn’t define a
single type of EE but rather allows for a variety of them. For example, complex services
in the control or management plane may be well executed in a user-space EE providing
considerable flexibility and robustness while services in the data plane may require a
kernel-space EE for performance reasons.
We want to mention two types of EEs developed in the FAIN project, namely the
JAVA-based EE [4] and the PromethOS EE [5]. With the JAVA EE complex services can
be rapidly prototyped making use of the broad functionality provided by the JAVA API.
It is also used for the implementation of the FAIN active node’s management layer. The
PromethOS EE is an extension of the Linux Netfilter framework and running in kernel-
space. In particular it allows to dynamically add services in the data path in the form of
modules. Thus it is predestined for services working on data packets and requiring high-
performance which wouldn’t be possible when data had to be copied from kernel-space
to user-space and vice versa.
For specifying, implementing, and managing services a component-based approach
was chosen. This approach offers a couple of advantages:
Aspects such as lifecycle management, configuration, access control, monitoring,
etc. can be separated from the service logic and are supported by the components’
runtime environment. Thus the developer of a service can concentrate on the service
logic and frequently needed aspects don’t have to be implemented over and over
again.
Complex services can be built by combining available components.
By providing means to dynamically interconnect components a high degree of flex-
ibility can be reached. After the initial setup services may be reconfigured during
runtime by adding new components, removing components, or changing connec-
tions between components.
Other systems with different or no component abstractions can be wrapped with proxy
components residing in the management layer thus making available the mentioned
advantages. Similarly, the abstraction of modules provided by the high-performance
PromethOS EE was mapped to components on the management layer.
The FAIN Active Node architecture defines some basic node services which are
available to later installed services. The basic services comprise management of com-
ponent environments, control of access to services and resources, packet dispatching
among components, and traffic control.
There are two different kinds of component environments: the already mentioned
execution environments (EEs) implemented with a specific technology and virtual en-
vironments (VEs) which abstract from a particular EE implementation and are used to
group resources belonging to the owners of VEs. A special privileged VE is owned by
the node’s operator and has full access to the node’s resources. This is where the basic
node services are installed when the node is booted.
137
The next section will describe a specific service used for controlling the access to a
WLAN. The service is made up from components in the data and control planes and is
running in two different types of execution environments.
4 Wireless Access
Based on the component model of the FAIN project a service for controlling access to
a WLAN, that takes into account the requirements posed by scenarios as discussed in
section 2, has been implemented and demonstrated during the FAIN project [6]. The
service effects load balancing between a number of wireless access points. In order
to provide support not only for this particular service, but for a variety of services a
key feature in this application domain – a generic framework for access control has been
developed. This framework makes use of the advantages of the kernel-space PromethOS
EE for processing data packets and the user-space JAVA EE for control purposes. With
this framework one can build access controllers equivalent to commercially available
solutions. Moreover, by exploiting the flexibility provided by the EEs it is possible to
deploy new services on-the-fly and to customise their behaviour according to different
applications’ needs. The establishment of application dependent resource control and
virtual network environments for the provision of resource shares to different users are
the key features of our implementation.
The framework for access control to wireless networks defines three entities situated
at different locations:
terminal daemons located at the end users’ terminals,
a WLAN access controller (WAC) located on a potential active node in the access
network, and
components in the data plane on one or more active nodes supporting the WAC.
A sample network setup showing the deployment of the different components is
presented in figure 1. The details of their functionality are explained in the subsequent
paragraphs.
Terminal Daemon: The terminal daemon is responsible for managing the WLAN
driver of the terminal. It determines the access points to be used for association and
controls the WLAN roaming behaviour. Further, it obtains information about the avail-
able access points and their signal quality from the WLAN driver. The terminal daemon
sends this information to the WAC and receives instructions which are then enforced us-
ing the terminal’s WLAN driver.
WLAN Access Controller: The WAC’s main task is to exercise control on all issues
related with the management of users and usage of the WLAN. Based on the informa-
tion received by the terminals and its own monitoring results, it decides on access and
load distribution in the network. Configuring the FAIN active nodes is a further task
of the WAC. The WLAN Access Controller is implemented based on the JAVA EE of
FAIN.
138
Terminal/
Terminal Daemon
Router
Access
Point
Access
Point
Remote
Server
WLAN Access
Controller
Terminal/
Terminal Daemon
PromethOS
Active
Node
Module_1
Module_2
Fig. 1. Deployment of WAC system components
PromethOS Modules: The components in the data plane EE i.e. modules in the
PromethOS EE are responsible for IP layer decisions concerning the traffic in the
WLAN network. They are configured by the WAC. In this respect the functionality of
the network node may be freely configured by the deployment of PromethOS modules.
However, a certain minimal set of basic functionality is always required:
routing/bridging between different subnets of WLAN access points and other net-
works,
load monitoring of data flows from and to access points per user or by other moni-
toring criteria,
routing to simulate bridging functionality of the active node, and
receiving messages destined for the WAC.
An example demonstrating the interaction of the different entities for a load balanc-
ing service is shown in Figure 2. It shows a handover between different WLAN access
points. The terminal daemons are constantly doing measurements about the available
access points and their signal strengths. This measurements are reported to the WAC.
Similarly, the traffic through the active node is measured and results are reported to the
WAC. Based on this information the WAC decides which terminal should be assigned
to which access point. In the case that a terminal should switch to another access point
the WAC sends instructions to the corresponding terminal daemon.
Based on the FAIN component model a very rapid development of the different
components could be made. In particular the JAVA components can be very efficiently
programmed. The PromethOS EE components are more sophisticated to program, as
they are implemented as Linux kernel modules.
Depending on the various needs of applications running on mobile terminals differ-
ent services can be deployed on demand. Making use of the component-based approach
a new service for wireless access control is deployed by installation of modules for the
139
Perform Handover
Handover-Command
WLAN Access Controller
Handover-
Command
Handover-Command
Measurements
Measurements
Measurements
Access Point
Access Point
Fig. 2. Interaction of entities in a handover scenario
PromethOS EE and WAC components for the user-space EE. The installation of service
components to different EEs is coordinated by the VE to which the EEs are attached.
The VE concept with the separation of resources allows to install multiple services in
parallel sharing the same underlying infrastructure.
5 Conclusions
In this paper, we presented the FAIN architecture for managing active nodes and an
application based on it for wireless networks. The introduction of virtual environments
in FAIN allowed integrating several execution environments with potential different
implementation technologies. Further, physical resources could be partitioned among
several node users the service providers with the help of virtual environments. To
achieve a flexible and fine grained control over service deployment and management
a component-based approach was chosen for the node level management layer. The
application in a wireless access systems turned out to offer a considerably increased
flexibility compared to commercial off-the-shelf solutions already available for access
control in WLAN networks. Another strength of the Active Networking concept is the
possibility to effect an application dependent behaviour of network elements. The po-
tential benefits of these capabilities that are easily installed in the framework for wire-
less access presented here remain to be investigated and are a field of further study.
Acknowledgement
This paper describes work undertaken in the context of the FAIN project (www.ist-
fain.org). We would like to thank our colleagues of the FAIN consortium for valuable
140
discussions, in particular Lukas Ruf from ETH Zurich for implementing and supporting
the PromethOS EE. The FAIN project was partially funded by the Commission of the
European Union as IST project 10561.
References
1. Bhattacharjee, S.: Active networks: Architectures, Composition, and Applications. Ph.D. The-
sis. Georgia Tech (1999)
2. IEEE Standard 802.11, Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) specifications (1999).
3. Kassler, A., Schorr, A.: Generic QoS aware Media Stream Transcoding and Adaptation. In:
13th Packet Video Workshop. Nantes, France (2003)
4. A. Galis, S. Denazis, C. Brou, and C. Klein: Programmable IP Networks: Management and
Rapid Service Deployment, Artech House London (2004).
5. Keller, R., Ruf, L., Guindehi, A., Plattner, B.: PromethOS: A Dynamically Extensible Router
Architecture Supporting Explicit Routing. In: Proceedings of the Fourth Annual International
Working Conference on Active Networks (IWAN 2002). Lecture Notes in Computer Science,
Vol. 2546. Springer Verlag, Berlin Heidelberg New York (2002)
6. Flury, P. (ed.): FAIN Demonstrators. FAIN Public Deliverable 40 (2003). http://www.ist-
fain.org/deliverables/del40/d40.pdf
141