Enabling Adaptive Service Access Management for
Next Generation Multi-Service Networks
Monique Calisti and Dominic Greenwood
Whitestein Technologies AG
CH-8032, Zurich, Switzerland
Abstract—This paper describes the Living Systems Adap-
tive Service Management, LS/ASM, Suite, an innovative and
comprehensive middleware solution that assists operators with
effectively delivering next-generation converged services by dy-
namically combining end-users and service-centric requirements
with network-facing management and control functionality.
The main goal is to discuss how a comprehensive policy-driven
and autonomic solution such as the LS/ASM Suite, spanning
basic infrastructures and end-user devices and building adaptive
control functionality directly into the corresponding elements,
enables the shift of focus from technology to value-addedservices.
The pervasive use of IP and the increasing availability of
ubiquitous broadband access in combination with advanced
wireless technologies are making network convergence a re-
ality. In this way operators aim to empower Telecom end-
users with the means to access a growing mix of value-added
services and applications that are made available to them at any
time, from any location and using any available access tech-
nology. This demands a common multi-access aware service
delivery platform able to seamlessly offer a reliable, secure,
easy-to-use and personalized service experience.
With this objective in mind, operators are expected to
migrate towards a new kind of service management approach
since most of the traditional client/server solutions turn out
not to be neither very effective or entirely appropriate due to a
lack of ability to handle the increasing dynamicity and diver-
sity of multi-service networks. In this perspective, emerging
solutions need to become increasingly “autonomic”, meaning
that their components should be able to self-conﬁgure thereby
dynamically optimizing their own operations according to the
way their environment and usage model changes .
LS/ASM, the Living Systems Adaptive Service Manage-
ment Suite, is a comprehensive and innovative solution that
enables effective delivery of next-generation converged ser-
vices by dynamically combining end user requirements and
service provisioning policies with network-facingmanagement
and control functionality.
Published in the Proceedings of the 4th European Conference
on Universal Multiservice Networks (ECUMN’2007), Febru-
ary 14-16, 2007, Toulouse, France.
By automating selected low-level processes on both the
users and operators sides and bringing more “personal intel-
ligence” - users context and behavior awareness - and “net-
work intelligence” - network services, content and resources
awareness - throughout the whole service delivery chain, the
LS/ASM solution realizes Adaptive Service Access Manage-
ment, ASAM. The central idea behind the ASAM vision is to
use autonomic techniques that enable operators to efﬁciently
manage and optimize resource utilization, performance and
end user experience. This is achieved by transparently tuning
service parameters while taking into account changes in both
the client and network context.
This paper ﬁrst introduces the complex multi-service net-
work context and highlights the main challenges associated
with effective management of services and resources. The
Adaptive Service Access Management, ASAM, approach is
proposed as a means to address these challenges by integrating
a set of features contributing to a more efﬁcient and ﬂexible
usage of network resources and delivery of services. The
LS/ASM Suite is then described in terms of its conceptual
foundation as a means to realise the ASAM vision, its archi-
tectural design, key features and beneﬁts, and several example
II. A CHALLENGING CONTEXT
In today’s competitive and convergingTelecom markets, the
increasingly ubiquitous and heterogeneous communication and
computing environments offer attractive business opportunities
to operators, but also pose signiﬁcant new challenges in many
areas of network and service management.
Telecom users are much more demanding. They request
new services to support a seamless and consistent experience
across multiple access technologies, devices and locations.
They expect to be always best-connected, i.e., they expect
anywhere and anytime access to the best available technology
with the maximum capacity on offer, plus easy-to-use and
problem-free services, all at ever lower prices.
The availability of new advanced end users devices enables
a variety of multimodal deployment scenarios, as depicted in
Figure 1, but also poses signiﬁcant challenges in terms of ser-
vice usability and personalization. In parallel, the widespread
proliferation of multiple broadband access technologies such
as cable, DSL, powerline, satellite, and wireless, is facilitating
3G Cellular Network
3rd Party Application Servers
Video on demand
Personal Mobile Services:
Location Based Services
Entertainment (TV, etc.)
Medical Mobile Services:
Access to patient records
data exchange (accident
site and hospital)
Fig. 1. A complex multi-service network scenario example.
the entry of new service providers in boththe ﬁxed and mobile
Therefore, to attract and retain customers, todays operators
need a cost-effective means to deliver value-added and reliable
services in a quick and ﬂexible way. In this perspective,
they need to make the best use of their current technology
investments by optimizing the way network performance and
availability are managed.
This is not an easy task, especially when considering that
many operators must deal with a diverse mix of systems and
processes that make it difﬁcult to effectively monitor and tune
services performance once delivered.
A. Adaptive Service Access Management
Because of the increasing deployment of multiple access
technologies at the edges of networks, the management of
next-generation services is changing rapidly. Intelligence and
speciﬁc management functions need to be migrated towards
the edge of the network and even onto the customers’
In particular, the set of functions including the selection
and maintenance of one of several available communication
channels and the adaptation of services taking into account a
variety of criteria and changes in the environment is indicated
as service access management. The major problems associated
with this process are:
•The fast and appropriate adjustment of the relevant con-
nectivity parameters to a continuously changing network-
•The assurance of sufﬁcient service quality and reliability,
whose perception can vary from one person to another
and thus needs to be addressed individually.
•In balance with the aforementioned points, the optimisa-
tion of resource usage and reduction of operational costs.
Adaptive Service Access Management, ASAM, addresses these
issues by providing the means to dynamically adapt the con-
ﬁguration and usage of available network access resources in a
reliable and cost-efﬁcient way. This is achieved by embedding
specialized “intelligence” into complex multi-technology and
multi-service access networks, including end user devices.
The central concept is to deploy smart techniques allowing
operators to efﬁciently manage and optimize resource utiliza-
tion, performance and end user experience. This by trans-
parently tuning service parameters (e.g., bandwidth, average
delay), while taking into account changes in the context,
including user proﬁle/preferences, Service Level Agreements,
SLAs, user location, terminal/devices features, and network
ASAM bases its adaptivity on the capability to au-
tonomously observe, extract, understand and use context infor-
mation to consequently modify its functionality. Information
exchange and correlation between client devices and access
nodes, as well as between access nodes even of different tech-
nologies, is at the core of this approach. In particular, through
dynamic mediation between (often conﬂicting) requirements
on the client and network side, capacity for given connections
requests is allocated by taking into account the status of the
whole service provisioning chain. This requires accounting
for a variety of parameters characterizing the connection to
be created and consequent required network resources, and
existing policies both on the user and provider side.
For this to be realized, ﬂexible and distributed monitoring,
conﬁguration and maintenance tools need to be smoothly
interfaced and integrated within the evolving networking en-
vironment and pre-existing management systems.
B. A New Solution Foundation
The strong limitations of traditional solutions for adaptive
service management and thereby delivery are directly related
to the lack of efﬁcient methods and tools capable of ﬂexibly
and dynamically mediating between providers needs and end
users requirements, while proactively monitoring and conﬁgur-
ing required resources. This is compounded by the difﬁculties
encountered by human operators when attempting to consider
all factors that inﬂuence the service management process and
make decisions in real-time.
In this perspective, a new kind of management solution
is needed. A comprehensive policy-driven and autonomic
architecture, spanning basic infrastructures and end-user de-
vices that builds adaptive control functionality directly into
the corresponding elements, enabling the shift of focus from
technology to value-added services. This is very complex also
because of the diversity of end user devices, network premises
and legacy systems such a solution should be integrated with.
The LS/ASM Suite is an innovative ASAM solution that
addresses these challenges by making use of multi-agent
technology concepts and tools  in combination with pow-
erful resource allocation algorithms and reasoning strategies.
Autonomous software agents that adapt to changes in the
environment, minimizing human intervention and service in-
terruption provide a powerful mean to engineer a distributed
and autonomic system that includes:
•Customizable and adaptive routines for automating and
tuning repetitive information and control tasks.
•Coordination mechanisms enabling dynamic collabora-
tion and aggregation of services.
•Abstraction of communication components to support
context changes through adaptation of semantic ground-
The central idea is that loosely-coupled distributed manage-
ment functions and control methods can be well-modeled and
implemented by making use of automated, goal-driven and
proactive software entities. These lightweight components are
able to operate on resource-scarce devices and support asyn-
chronous communication with intermittent network connec-
tions. Moreover, according to the results of proactive monitor-
ing information received from the environment within which
they are embedded, the LS/ASM components directly assist
with autonomic management of network resources. They are
able to conﬁgure themselves and dynamically optimize their
operations according to the way their environment changes
and in-line with operator and client user policies. They thus
assist with the speed-up and automation of simple, tedious and
repetitive service management tasks currently performed most
commonly by human operators. The ultimate result of this is
potentially substantial cost savings to the operator.
C. Related Work
Although other approaches     have been
proposed in the literature that address part of the ASAM
challenges, few, to our knowledge, deal with the speciﬁc case
of dynamic mediation between network and client require-
ments and accommodate resource allocation and consumption
In particular, the solution presented in  deﬁnes support
for vertical handover in Radio Access Networks with selec-
tion criteria predominantly based on Quality of Service. A
vertical handover decision module is described in terms of its
placement and operation within a concrete provider system.
This decision module can communicate with various network
devices, including client devices, to determine radio access
network selection based on QoS parameters. Some degree of
negotiation takes place, but only between entities within the
network and excluding the clients. The client devices remain
The work described in  proposes a market-based ar-
chitecture where software agents are used to negotiate the
provisioning of a service by service providersto client devices.
This article describes only a single, non-iterative, negotiation
phase triggered by location change between a user device and
service providers. It also has no account for cooperative offers
created by multiple contributing network provider devices that
have derived a proposal by negotiating amongst themselves.
This approach also only considers providers that offer WLAN
connectivity, rather than multiple possible wireless connection
types (including cellular and WLAN technologies). All of
these points are signiﬁcant differences in comparison with the
Other related approaches are discussed in Section V when
discussing in more details the deployment of the LS/ASM
solution in speciﬁc scenarios.
III. THE LS/ASM SUITE
As depicted in Figure 2, the LS/ASM architecture includes
two main types of autonomic software components, which
communicate by relying upon the use of common interaction
User Data, Control Data,
LS/SAM Logic: AAM
LS/CA Logic: ALM
Fig. 2. The LS/ASM architecture.
protocols and a shared semantics-based ontology deﬁning all
LS/ASM needed concepts.
•The Living Systems Connection Agent, LS/CA1, is a client
component that can run on a variety of mobile end user
devices (e.g., laptops, PDAs, smart phones) and provides
mobile users with improved quality and reliability by
optimizing service access through adaptive connection
handover across multiple access technologies. By taking
into account user preferences, application requirements,
devices properties and dynamic network conditions, the
LS/CA manages the end user device’s connection to the
•The Living Systems Service Access Manager, LS/SAM,
is a network component that can run on hardware lo-
cated at the access nodes2or at a network management
facility and dynamically manages and optimizes resource
allocation using differentiated access technologies with
adaptive problem recovery and load balancing methods.
Proper actions (e.g. trigger vertical handover, reconﬁgure
a device) are performed by the LS/ASM according to
operators policies and well-proven distributed resource
These lightweight software components can ﬂexibly com-
plement and extend a number of existing service management
architectures, and are able to run on resource-limited devices
and support asynchronous communication with intermittent
network connections. In addition, by dynamically coordinating
their actions and behaviour, they enable adaptive communica-
tion service access by mediating between operator policies and
end-users requirements and preferences.
This is achieved through a negotiation procedure that can be
initiated either by a LS/CA calling one or several LS/SAMs
for a proposal to set up a connection accordingto speciﬁc end
user’s requirements, or by a LS/SAM sending a connectivity
offer to a LS/CA based upon current network conditions and
1LS/CA is available as a standalone commercial product (see
2Access nodes typically are send/receive stations that provide a link to an
operator’s policies. The connection characteristics on which
negotiation is based can comprise bandwidth (average, mini-
mum and maximum), maximum end-to-end delay, acceptable
jitter, network type and cost schedule.
A. The LS/ASM Client Component
The LS/CA includes the Adaptive Logic Module, ALM,
which provides adaptive service access by setting the net-
working parameters accordingto the outcome of the mediation
process between the end user’s requirements and the network
provider’s offering (as speciﬁed by the involved LS/SAM).
The ALM outcomes and decisions are determined by a set of
•Quality requirements of the applications and services
running on the device the LS/CA is embedded in.
•Physical end user device status, e.g., battery power level
that can affect the transmission technology selection, and
properties, e.g., only speciﬁc access technologies might
•Existing service provisioning conditions according to pre-
deﬁned subscription contracts/service level agreements.
The ALM proactively manages and processes this informa-
tion according to policies which capture end user preferences,
e.g., minimising connection costs, maximising battery life
when on-the-move, etc. These policies are ﬂexible in that they
may be adapted automatically or changed by the user when
B. The LS/ASM Network Component
The LS/SAM includes the Adaptive Access Module, AAM,
that proactively monitors trafﬁc and resources in the access
node the LS/SAM is controlling, triggers appropriate actions
(e.g., vertical handover, load balancing, congestion recovery)
according to the network status and current trafﬁc conditions,
processes incoming LS/CA calls for proposal and thereby
elaborate offers as appropriate.
The AMM decisions and behaviour are guided by the oper-
ator’s policies that express provider preferences with respect
to a variety of aspects including, for instance, how to allocate
trafﬁc to balance out network utilization, how to treat speciﬁc
users (i.e., connection) in case of congestion, how to adapt
pricing schemes according to the user’s subscription type. This
requires dynamic monitoring and processing of information
•Trafﬁc conditions and resources availability within the
access node the LS/SAM is controlling.
•Trafﬁc conditions and resources availability in other ac-
cess nodes a given portion of trafﬁc can be handed over
to via dynamic peer-to-peer LS/SAMs coordination.
•Existing service provisioning conditions according to pre-
deﬁned subscription contracts/service level agreements.
C. Client-Network Mediation
The mediation process conducted between the LS/CA and
LS/SAM components consists of a sequential interchange
formulated as a contract-net protocol  negotiation with the
call for proposals
initiate service set up
[proposal = accepted OR iteration = maxallowed]
Loop for re-
Fig. 3. Mediation process between the LS/ASM client and network
goal of determining the best connection parameters given the
requirements of the end user, the offering of the network
provider and the conditions of the transmission medium.
The requirements of the end user toward the provider are a
combination of (i) the preferences of the end user formulated
as user policies (e.g., minimising connection cost), (ii) the
quality demands of the applications running on the end user
device (e.g., a given application may require low end-to-end
delay), (iii) the status of end user device resources (e.g.,battery
power, which can affect the selection of the transmission tech-
nology), (iv) the technologies supportedby the end user device
(e.g., only WLAN and UMTS network interfaces available),
and (v) the conditions stated in the subscription contract (e.g.,
costs for using certain technologies).
The offering of the provider toward the end user is de-
termined by considering (i) the properties of the provider
network (e.g., diversity of network access technologies), (ii)
the network status (e.g., distribution of trafﬁc load, delay
times), (iii) the capabilities of the network (e.g., mobility
support, QoS control) and (iv) the provider policies, including
business rules, that relate to the use of its infrastructure, pricing
schemes, trafﬁc prioritization mechanisms, etc.
Figure 3 illustrates the typical message exchange during a
proposal setup sequence. The LS/CA sends a Call For Proposal
(CFP) to one or several LS/SAMs requesting offers to set
up a connection with speciﬁed constraints including quality
requirements, or connection characteristics.
A simple example CFP is as follows:
CFP: (set up connection, (min. bandwidth: 100 Kbit/s, max.
jitter: 50 ms))
In addition, speciﬁc time schedule information regarding the
start time and end time of the connection might be included
(e.g., start-time: 2006-09-01, 12:00 UCT, end-time: 2006-09-
01, 16:00 UCT). In special cases, where the connection needs
to be periodically established, the recurrence of this event can
also be stated in the proposal.
The proposal offered by the involved LS/SAM is then
determined as a function of the supplied client constraints on
what a proposal may contain, provider policies and network
state, which may relate to the instantaneous state and/or his-
torical state data, optionally including reservations of network
An example of a proposal is as follows:
Proposal: (set up connection, (network: UMTS, min. band-
width: 100 Kbit/s, max. bandwidth: 120 Kbit/s, max. jitter: 40
ms, max. end-to-end delay: 200 ms))
This example proposal lists additional parameters than those
mentioned in the original CFP. These can be taken into account
by the receiving LS/CA, or ignored as preferred.
The originating LS/CA waits a predeﬁned duration to re-
ceive incoming proposals and/or rejections. Once the deadline
is reached, the LS/CA begins to assess received proposals
by considering (i) the set of quality requirements stated in
the original CFP, (ii) the received proposal (or the relevant
parameters stated in the proposal), (iii) the user policies, and
(iv) the status of the end user device. The output of the
assessment function is a value indicating the viability of the
proposal. If no proposals are acceptable, the CFP can be
revised and resent to the same or different LS/SAMs. This
forms the basis of the mediation process which may take place
over several iterations. A basic protection algorithm is used to
ensure that negotiation is convergent,thereby avoiding lengthy
or endless iterations.
When, or if, a proposal is accepted the client device sends
an accept-proposal message to the corresponding network
provider. All other proposals that have been received are
explicitly rejected by informing their source providers. The
reason for rejection may be included in the message.
IV. REALIZING ASAM AS LS/ASM
The dynamic coordination of LS/ASM components en-
ables mobile data connectivity and service management to
be optimised transparently across multiple network access
technologies. In particular, the ASAM vision is realized by
means of a comprehensive set of mechanisms including:
•Seamless handover and session continuity that guarantees
interruption-free service access across multiple technolo-
gies by allowing an LS/CA empowered device to main-
tain the same IP address for an entire session. This is
achieved by making use of Mobile IP technology 
and (alternatively) SESAM technology .
•Secure communication. Tight integration of the LS/CA
with several 3rd party VPN clients allows an always se-
cure connectivity. Furthermore, by integrating IPSec 
and Mobile IP, the LS/CA ensures end-to-end encryption
of all generated trafﬁc (as an optional feature).
•Connection adaptation, which means automatic detection
of available networks and selection of the preferred
network adapter (access technology) based on service
requirements and network conditions for improved relia-
bility and QoS. This, as explained in the previous section,
can trigger dynamic negotiation between the LS/CA and
the LS/SAM components.
•Context-aware customer support through semantic ser-
vice speciﬁcations, policy-driven decision making and
dynamic information retrieval. The LS/CA improves end-
user experience by directly dealing and addressing low-
level issues (e.g., failure recovery, connection adaptation),
while taking into account user policies and boundary con-
ditions, i.e., context-based information and coordination
with LS/SAM components as needed.
•Congestion recovery: instantaneous and proactive detec-
tion, analysis and relief of congestion, which reduces
call dropping and increases services resilience and avail-
ability. Within an access node, once no new network
connection can be accepted or the total requested band-
width exceeds the total available one, i.e., packets are
dropped, an LS/SAM can decide upon speciﬁc strategies
and existing service level agreements (if any) whetherand
how to drop or hand over part of the trafﬁc to neighbour
•Load-balancing: balance trafﬁc load across WiFi and
cellular networks while considering the QoS needs of
running services, making the network more resilient to
trafﬁc peaks. This is achieved by dynamic coordination
of LS/SAMs that can hand over a certain number of con-
nections to neighbour access nodes according to several
possible operator strategies. The use of distributed con-
straint satisfaction algorithms  enables to effectively
balance out the load by taking into account all existing
hard and soft constraints.
A. Beneﬁts of LS/ASM
By means of the features described above the LS/ASM Suite
creates value by bringing important beneﬁts to both end users
LS/CA elements provide end users with the best connection
available in wherever they may be located, e.g., an train,
airport, business meeting, or at home. They achieve this by
enabling a terminal to seamlessly select the access technology
that will provide the best connection (e.g., that which offers
the highest data rate), while automatically taking into account
user requirements and provider policies.
By using proven distributed resource allocation algorithms
for congestion recovery and load balancing, coordinating
LS/SAM entities are able to optimize the deployment of net-
work resources. Furthermore, resource optimization strategies
are applied not only locally, but also take into account the
status of the whole provisioning chain (through peer-to-peer
LS/SAMs coordination mechanisms3). Moreover, historical
data relating to the provisioning chain can be aggregated and
used for off-line analysis to support network dimensioning and
LS/ASM optimization techniques can also take into account
different types of user subscriptions, by enabling service
differentiation and premium services both per-session and
per-subscription. This enables operators to adopt an effective
convergence strategy taking better advantage of the unique
beneﬁts provided by each deployed access technology, while
better serving customers with constantly growing demands.
3Discussion of peer-to-peer LS/SAM coordination, both intra- and inter-
operator, is omitted from this paper due to space limitations.
In addition, the ﬂexible nature of deployment using soft-
ware agents allows straightforward support for fault tolerance
through the integrated use of autonomic self-management
mechanisms. In the ﬁrst instance, if a negotiated connection
fails because an LS/SAM becomes unavailable, it can be
quickly replaced with an alternative connection established
through an earlier, or on-demand, negotiation with alternative
LS/SAMs. Secondly, both LS/CA and LS/SAM agents can,
if necessary, be hot-swapped with a ready-to-run redundant
backup agent that copies transaction controlled state and takes
over if a failure occurs in the primary.
V. LS/ASM DEPLOYMENT SCENARIOS
Different combinations of the LS/ASM features enable a
variety of deployment scenarios. In the following, three of the
most signiﬁcant ones are presented including a discussion of
the distinctive characteristics in relation to relevant work.
A. Enforcing QoS
The notion of guaranteed data transmission quality with
enforcement mechanisms, in particular for emerging QoS
sensitive multimedia applications, e.g., voice or video over
IP, is a key issue especially in converged networks . While
trafﬁc prioritization is often not of paramount importance in
core networks due to overprovisioning, QoS is an essential
differentiator in limited-capacity wireless access networks for
capacity and/or delay sensitive trafﬁc such as voice or video
over IP. While for cellular access technologies belonging to
2.5G, 3G and 3.5G, appropriate standards for QoS have been
deﬁned, few operators yet make widespread use of them. In
addition, the WiFi world is supporting its technologies with
speciﬁcations that directly account for QoS management.
In particular, when integrating different access network
technologies, e.g., WiFi and UMTS, the quality of a connec-
tion may be degraded during vertical handover where (i) the
connection needs to be re-established at the new access node,
which is time consuming and during which no data can be
transmitted, and (ii) if too many IP packets are lost, they must
be retransmitted which can also be time consuming in the
case of a large number of packets - again leading to service
Various approaches have been developed and proposed
to address this problem. In , a reservation-based QoS
model for integrated cellular and WiFi networks is deﬁned
and an adaptive mechanism to ensure end-to-end QoS is
proposed. However, this model can only work by making
the assumption that cellular/WiFi interworking is realized by
relying upon a common and uniform reservation-based QoS
architecture, which is not (yet) the case for most real network
scenarios. Similarly, Song et al.  proposed an admission
control mechanism for integrated voice and data services in
cellular/WiFi networks. The main limitation of this approach
though is that it does not account for video trafﬁc.
To effectively provision QoS and optimize resource uti-
lization for a variety of possible heterogeneous network sce-
narios, the LS/ASM Suite relies upon the dynamic combi-
nation of speciﬁc mechanisms both at the client side (i.e.,
UMTS base station
WLAN access point
No specific QoS
Fig. 4. Deployment model of the LS/ASM Suite for QoS enforcement.
seamless handover, session continuity and connection adap-
tation) and at the network side (i.e., congestion recovery and
load-balancing) that are compliant with dominant industrial
standards, e.g., mobile IP or SIP/IMS, when supported, or
technology-independent, whenever possible.
Unlike legacy systems and hardware-based solutions, the
LS/ASM software components accommodate high-level ser-
vice and user needs and preferences (including QoS re-
quirements) by implementing coordination mechanisms and
resource allocation algorithms that hide low-level access tech-
nology dependent processes. This is achieved by deploying an
agent-based middleware architecture that provides users with
a common and higher level of abstraction, which makes low-
level network access heterogeneity transparent.
On the client side, basic QoS in terms of service avail-
ability and continuity is enforced by the LS/CA through
automatic and policy-driven vertical handover, i.e., all trafﬁc
is switched from one network interface, according to existing
constraints and user policies. Moreover, by continuouslymoni-
toring network conditions and device status and properties, the
LS/CA exerts QoS and context-aware resource management
by selecting the most appropriate access technology to be
used for the running applications/processes. In addition, when
appropriate, as detailed in Section III-C, the LS/CA can also
trigger negotiation with one or more LS/SAMs for different
On the network side, the key mechanisms deployed by
the LS/SAM to enforce QoS provisioning are load-balancing
and congestion recovery. Load-balancing can be triggered by
LS/SAMs in order to redistribute trafﬁc across several access
nodes according to various criteria, including:
•Current utilization of resources at the access node, e.g.,
once the trafﬁc overcomes a given threshold a certain
portion of the supported connections might be handed
over to neighbor LS/SAMs.
•QoS requirements of the running services, e.g.,best-effort
connections might be handed over to prioritize premium
services for which charging might be based on service
reliability guarantees (e.g., ≥95% non-disruption).
•Predictions of the network resources usage to minimize
the probability of congesting an access node.
Analogously, whenever congestion occurs a speciﬁc part of
the trafﬁc at a given access node might be handed over to
other LS/SAMs or selected existing connections (e.g., the non-
premium ones) might even be dropped as appropriate. This
enables relief of congestion and increases service resiliency
For example, assume a user that launches an IP-based TV
program (e.g., a news channel) on a smart phone. During the
launch of the selected application to render the video stream,
the LS/CA determines the connectivity parameters (typically
bandwidth and delay) for interruption-free high quality service
provision. Because different access technologies offer different
QoS assurances, the LS/CA might trigger vertical handover to
a speciﬁc technology, e.g., UMTS, that better supports the
QoS level needed for the video down-streaming. In addition,
in the case of an UMTS connection, the LS/CA would set up a
new Packet Data Protocol context requesting the UMTS QoS
streaming class .
Figure 4 depicts the deployment model for this case. Each
end user device is installed with an LS/CA component able
to enforce QoS. The LS/CA must be aware of the different
trafﬁc categories available in each network access technology.
During a vertical handover,the QoS class of the active network
is mapped into an appropriate QoS class of the target network.
There is one LS/SAM agent being deployed per access node,
i.e., each LS/SAM agent is in charge of a speciﬁc access node
and thus is up-to-date at all times regarding the status of that
node. When planning load balancing and congestion recovery,
the LS/SAM agent must be aware of the QoS classes supported
by the different access technologies to minimize the risk of
degraded service quality.
B. Integration with Push Services
Push technologies  are widely used in today’s mobile
communications market to send media content to users as
soon as it becomes available. However, the majority of current
existing solutions are neither able to determine the most
suitable moment to send content nor which content quality
to use to deliver it.
The LS/ASM Suite can be integrated with third party push
service platforms to improve the media delivery process by
providing information about the current status of the target
user device and capability/reliability of the access network to
be used to better reach such device. In this way, it is possible
to more effectively choose when to push the content and at
which quality it should be sent (according to the available
As a simple use case to illustrate the combined deployment
of a generic push platform and the LS/ASM Suite, assume
a mobile user connected to a cellular network (for instance
UMTS) through which he/she is engaged in a video con-
ference. As depicted in Figure 5 the LS/CA component is
installed on the end user device. The LS/SAM component is
End User Device
Logical Integration Points
User device and
access node network
Fig. 5. Deployment model of the LS/ASM Suite for the push service scenario.
installed on a server that resides inside the operator’s network
- one LS/SAM component per each existing access node.
In addition, an Adaptive Push Agent (APA) component is
integrated with the push platform.
When passing a particular shopping outlet, with which
the user has pre-registered, the push platform is signalled to
transmit an advertisement to that user in either image or video
The APA contacts the LS/SAM controlling the access node
to which the user is connected in order to retrieve appropriate
information (e.g., capacity and security characteristics) about
the deployed network connection.Based on this, the APA ﬁnds
out that sending a video advertisement to the user would be
disruptive to his/her already running video stream and thus
also to the user’s quality of experience. Therefore, the APA
triggers the advertisement to be sent in image format with a
text tag indicating that a video advertisement is also available
if the user wishes to receive it in a later moment.
C. Integration with IMS/SIP
IP Multimedia Subsystem, IMS, initially developed by
3GPP and 3GPP2 as an IP core network architecture for
cellular/wireless-based access to Internet services, is now
evolving into a standard that provides a common frame-
work to create and offer next generation converged network
services . IMS builds on the Session Initiation Protocol,
SIP, that is mainly in responsible for delivering a session
description to a user at its current location . Thekey idea is
to enable any kind of access (wireless or ﬁxed) for any kind of
media (including any combination of voice, text, image and/or
video) supporting multiple devices and endpoints.
Because of the (at least initial) co-existence of IMS and non-
IMS applications, the costs associated with moving to a full
IMS-based network, and the inherent complexity of IMS (and
its several standards, interfaces and protocols) most service
providers and or operators are expected to migrate toward an
IMS service framework iteratively.
One of the core issues to be addressed for successful adop-
tion of IMS is the ability to face more aggressive bandwidth
and latency demands, which implies increased QoS manage-
ment and design capabilities on the bearer network . In
particular, IMS/SIP lacks trafﬁc management capabilities and
especially adaptive connectivity management and optimization
mechanisms that can be regarded as key components for
delivering ubiquitous quality-sensitive multimedia services.
In this perspective, the LS/ASM Suite complements an
IMS-based framework by ensuring the quality of delivered
services at the bearer network level through its adaptivity
mechanisms, leaving IMS/SIP to cope with call control and
service deployment issues. As depicted in Figure 6 the LS/CA
component directly interacts with the SIP client installed on
the end user device. In this way, the SIP client is able to obtain
information on the quality of the connection which is helpful
to determine, for instance, the appropriate codec to use, and
to request the LS/CA component to ensure a certain quality
level (in particular, when explicit QoS class enforcement is
enabled). On the network side, an LS/SAM agent integrates
with each access node and, by means of load balancing and
congestion recovery enables to provide a high level of service
A simple use case is when one considers the collaboration
between a SIP client and the LS/CA component to guarantee
a level of quality required by a user to perform a video
call. Upon launch of the SIP-based video calling application,
the SIP client assesses the connection quality by means
of the LS/CA component. The SIP client is aware of the
quality requirements imposed by the video call service that
are also variable according to the size and quality of the
video picture. The LS/CA component can, in collaboration
with the respective LS/SAMs, discover the quality offering at
alternative access nodes and, based on that decide whether a
handover to another access node needs to be triggered. Both
end devices that participate in the video call must also agree
on the codecs to be used for encoding and decoding the voice
and video data. The LS/CA component delivers the necessary
information to the SIP client to make its choice. Once the
video call is established and running, it is the LS/CA agent’s
responsibility, in cooperation with the active LS/SAM agent,
to preserve the quality of the connection and take appropriate
measures if tolerance thresholds are violated. Depending on
the mobility proﬁle of the user, but also on the evolution of the
network conditions, handoffsare unavoidable and thus need to
be well planned and efﬁciently executed to minimize quality
The LS/CA does not affect the SIP call itself nor in-
fringe any of the IMS/SIP standards. SIP is concerned with
controlling the call execution while LS/ASM takes care of
connectivity. LS/ASM is therefore complementary to IMS/SIP
and beneﬁts result even if only a small proportion of the entire
network infrastructure (namely the access part) and end user
devices are LS/ASM empowered.
VI. CONCLUSIONS AND FUTURE DIRECTIONS
The LS/ASM Suite is a distributed and resilient system
that exhibits high adaptivity to its network environment. This
IP Core Network
Other control messages
SIP Client 3rd Party
Application Servers OSA
CSCF: Call State Control Function
GGSN: Gateway GPRS Support Node
OSA: Open Service Access
RNC: Radio Network Controller
SGSN: Serving GPRS Support Node
UE: User Equipment
UTRAN: Universal Terrestrial Radio Access Network
Logical Integration Points
Fig. 6. Deployment model of the LS/ASM Suite when integrating with an
has been achieved by properly combiningmulti-agent systems
concepts and technology with powerful resource allocation
algorithms and reasoning strategies. By hiding low-level net-
working aspects that, especially in next generation multi-
service network scenarios, can continuously change due to end
users mobility, the LS/ASM middleware provides transparent
service access in heterogeneous networks and becomes an
essential autonomic complement to (bearer unaware) service
However, to achievethe potential of autonomic management
approaches in todays’ networks is not a straightforward task.
Migrating intelligence and complex management functions
towards the edge of the network reduces the degree of manual
intervention needed, but can also increase the complexity of
the management system itself. The network must indeed be
adaptable, but at the same time stable and controllable. There-
fore, populating the networking environment with autonomic
software components requires some additional conﬁguration
and monitoring capabilities.
In this sense, middleware technologies for highly dynamic
and heterogeneous networks are required to become able
to monitor and control the middleware itself, by integrating
with traditional quite static infrastructures often populated by
legacy systems and adapting to different operating systems and
connection technologies. This is a challenging task that still
requires extensive research and experimental work.
Our ongoing and future work includes more extensive
characterization of LS/ASM beneﬁts when adopting differ-
ent user and operators policies, network allocation strategies
and algorithms. This is done by simulating and analyzing
the LS/ASM Suite performance in a variety of networking
scenarios and consequently reﬁning the behavior of the var-
ious system components. In addition, by means of selected
testbed demonstrations and experiments we are assessing the
feasibility and complexity of integrating LS/ASM entities in
speciﬁc service delivery frameworks. As a matter of fact,while
the LS/CA has been already successfully deployed in a variety
of real-world scenarios, the adoption of the LS/SAM requires
some additional work given the wide assortment of existing
and upcoming service and network management architectures.
We would like to thank colleagues at Whitestein Technolo-
gies who signiﬁcantly contributed to deﬁne and develop the
LS/ASM Suite and concepts, in particular Thomas Lozza,
Roberto Ghizzioli, Martin Stangel, Oliver Hoefﬂeur, Oliver
Carl, Thomas Haas, Giosu`e Vitaglione and the whole LS/CA
 3GPP, “TS 23.107 v7.4.0: Quality of Service (QoS) concept and
architecture,” June 2006.
 ABIresearch, “IP Multimedia Subsystem Industry Survey Results,” 2005.
 K. Ahmavaara, H. Haverinen, and R. Pichna, “Integration of wireless
LAN and 3G wireless - Interworking Architecture between 3GPP and
WLAN systems,” IEEE Communications Magazine, vol. 11, no. 41, pp.
 E. Bircher and T. Braun, “An agent-based architecture for service
discovery and negotiation in wireless networks,” in Proceedings 2nd
International Conference on Wired/Wireless Internet Communications,
Frankfurt, Germany, Feb 2004.
 M. Cuevas, “Admission control and resource reservation for session-
based applications in next generation networks,” BT Technology Journal,
vol. 23, no. 02, 2005.
 M. Danzeisein and T. Braun, “Secure Mobile IP communication,” in
Wireless Local Networks Workshop at IEEE Local Computer Networks
Conference, Tampa, USA, Nov 2001.
 R. Ferrus, A. Gelonch, O. Sallent, and J. Perez-Romero, “Vertical Han-
dover Support in Coordinated Heterogeneous Radio Access Networks,”
in Proceedings 14th IST Mobile and Wireless Communications Summit,
Dresden, Germany, June 2005.
 Foundation for Intelligent Physical Agents, “FIPA Iterated Contract Net
Interaction Protocol Speciﬁcation,” Dec 2002.
 L. A. Guedes, P. C. Oliveira, L. F. Faina, and E. Cardozo, “An
agent-based approach for supporting quality of service in distributed
multimedia systems.” Computer Communications, vol. 21, no. 14, pp.
 T. Guenkova-Luy, A. Kassler, and D. Mandata, “End-to-end quality of
service coordination for mobile multimedia applications.” IEEE Journal
on Selected Areas in Communications - Advanced Mobility Management
and QoS Protocols for Wireless Internet, vol. 22, no. 5, pp. 889– 903,
 I. Herwano, J. Sachs, and R. Keller, “Provisioning and Performance
of Mobility-aware Personalized Push Services in Wireless Broadband
Hotspots,” Computer Networks, vol. 49, no. 03, 2005.
 N. R. Jennings, “Agent-Oriented Software Engineering,” Lecture Notes
in Computer Science, vol. 1647, pp. 1–7, 1999.
 S. Kent and R. Atkinson, “RFC 2401: Security Architecture for the
Internet Protocol,” Nov 1998.
 M. Kolbehdari, D. Lizotte, G. Shires, and S. Trevor, “Session Initiation
Protocol (SIP) Evolution in Converged Communications,” Intel Technol-
ogy Journal, vol. 10, no. 01, 2006.
 C. Perkins, “RFC 3220: IP Mobility Support for IPv4,” Jan. 2002.
 W. Song, H. Jiang, W. Zhuang, and X. Shen, “Resource management for
QoS support in cellular/WLAN interworking,” IEEE Network, vol. 19,
no. 5, 2005.
 J. Strassner, “Autonomic Networking - Theory and Practice,” in Pro-
ceedings 9th IFIP/IEEE International Symposium on Integrated Network
Management, Nice, France, May 2005.
 X. G. Wang, G. Min, J. E. Mellor, K. Al-Begain, and L. Guan, “An
adaptive QoS framework for integrated cellular and WLAN networks,”
Computer Networks, vol. 47, no. 2, 2005.
 M. Yokoo and K. Hirayama, “Algorithms for Distributed Constraint
Satisfaction: A Review,” Autonomous Agents and Multi-Agent Systems,
vol. 3, no. 2, pp. 185–207, 2000.