Autonomics in Wireless Network Management: Advances in Standards and Further Challenges Kostas Tsagkaris, Panagiotis Vlacheas, George Athanasiou, and Vera Stavroulaki, University of Piraeus Stanislav Filin and Hiroshi Harada, NICT Jens Gebert, Alcatel-Lucent Bell Labs Germany Markus Mueck, Intel Mobile Communications Abstract The increase of user traffic demand, the numerous and QoS-pretentious applications and services, as well as the evolving and emerging business models comprise prominent characteristics of the Future Internet (FI). They also comprise challenges that will necessitate significant alterations in the way that networks delivering FI services are managed, and which are insufficiently addressed by existing solutions. In this respect, autonomic systems exposing self-management and learning capabilities have appeared as the most viable direction for tackling the foreseen complexity and realizing the FI vision. At the same time wireless networking and its evolution will constitute an integral part of FI and thus, the adaptation and integration of the related exploratory work in autonomics into the existing wireless network management is a major challenge to address. The concept of Autonomic Wireless Network Management (AWNM) is the result of this integration. AWNM has been investigated by many research efforts, but also engaged working groups within major standardization bodies. This article will go through the most relevant related standardization activities and explain how autonomics-related ideas are integrated. In particular, 3GPP SON is considered as a rather near-time activity of the new trend in autonomic networking in the wireless domain; on the other hand, it is shown that IEEE DySPAN 1900.4/1900.4a and ETSI RRS Working Group 3 address more disruptive ways forward and are therefore targeting a mid- to longerterm time-frame. Extrapolating the inherent trends, the expected further evolution of the intelligently self-managed wireless communications framework is finally outlined.
he vision of Future Internet (FI) has created great expectations for the business players and a great area for research in the information, telecommunication and networking research community. The increase of user traffic demand, the numerous and QoS-pretentious applications and services, as well as the evolving and emerging business models, are envisaged characteristics of the FI. A key part of this vision lies in the evolution of the wireless networking in order to enable users to be always-on, always-connected in an ever growing environment; in particular, the number of network nodes in the field increases exponentially and furthermore the overall trend is to deploy a large number of heterogeneous systems. “Heterogeneity” is being addressed in terms of distinct Radio Access Technologies (RATs), such as Wireless Fidelity (WiFi), Worldwide Interoperability for Microwave Access (WiMAX), Long Term Evolution of 3rd Generation Partnership Project (3GPP LTE), etc., but also building on a single-RAT framework applying different, possibly overlapping cell-types (in particular 3GPP Femto-, Pico-,
IEEE Network • November/December 2011
Micro- and Macro-Cells). Advanced systems need to manage all available RATs through their respective Radio Access Networks (RANs), exploiting the liberation and liberalization of several portions of radio spectrum and also building on devices with reconfiguring capabilities. It is a common truth that the centralized management of such a diverse system becomes quasi-impossible and thus autonomic systems and associated self-management and learning features have appeared as one of the most viable directions for tackling the foreseen complexity and realizing the FI vision . As an example, management tasks are transferred to the various infrastructure and user device nodes, which exploit context information and centrally handled policies in order to perform localized decision making and flexible adaptation to an ever changing context. Therefore, a key challenge lies in the efficient combination of legacy wireless network management with adapted exploratory results in the field of autonomic systems  as illustrated in Fig. 1. The confluence of wireless network management and auto-
0890-8044/11/$25.00 © 2011 IEEE
Centrally controlled legacy technology (2G, 3G, WiFi, etc.)
Next generation selfmanaged framework (LTE-advanced, etc.)
Wireless network management evolution
Autonomic/cognitive wireless network management
Autonomics concepts evolution
Figure 1. Evolution of wireless network management and autonomics concepts toward an integrated next generation framework. nomics is the so called Autonomic Wireless Network Management (AWNM). AWNM has been investigated by many research efforts , but at the same time it is the subject of working groups within major standardization bodies, namely, the European Telecommunications Standards Institute (ETSI), the Institute of Electrical and Electronics Engineers (IEEE), and 3GPP, since industrial players express a need for an immediate availability of related solutions. Framed within this statement this article presents an overview of recent advances in AWNM-related standardization, where such an approach is sought for an industrial and product-driven perspective. In particular, it covers ETSI Reconfigurable Radio Systems (RRS) Working Group 3 , IEEE Dynamic SPectrum Access Network (DySPAN) 1900.4/1900.4a [5, 6], and 3GPP Self-Organizing Networks (SON) , as being illustrative of the new trend in autonomic management in the wireless domain, advocating at the same time for the unique role that these groups are expected to play in the constitution of the self-managing FI. While 3GPP SON defines the self-management approach to be applied in the immediate future, ETSI RRS and IEEE DySPAN are addressing more advanced, fully heterogeneous systems-based approaches, which are expected to be deployed in the mid-term. It must be noted that in both ETSI RRS and IEEE DySPAN the term “cognitive” is interchangeably used with that of autonomic, thus referring to Autonomic/Cognitive Wireless Network Management standards, and it describes both the ability to operate and manage different spectrum bands in a dynamic and flexible manner and an advanced, intelligent processing and adaptive decision making ability, e.g. based on learning and knowledge diffusion. The rest of this article is organized as follows. The next section gives an overview and analyzes the above mentioned AWNM standards. Then we discuss the challenges and suggestions toward the future of AWNM. The article is finally concluded.
Overview and Analysis Of Autonomic/Cognitive Wireless Network Management Standards This section gives an overview of the standardization outcomes of ETSI RRS WG3, IEEE DySPAN 1900.4/1900.4a and 3GPP SON, respectively. ETSI RRS and IEEE DySPAN both cover mid-term dynamic spectrum management aspects in a single- or multi-operator, fully heterogeneous wireless network context. They introduce architectures that comprise a set of system/functional entities (managers) in network and terminal sides and distribute (self-)management tasks among them for realizing certain goals related to spectrum and resource usage optimization. 3GPP SON, on the other hand, targets immediate production and thus builds on a less disruptive approach: mainly 3GPP-RAT centric self-organization
tasks are defined to alleviate the drawbacks and complexity of deployment and operation of legacy cellular networks. The standards are analyzed by considering the way they abide by a set of important features as key building blocks of an autonomic/cognitive management cycle, namely context awareness, decision making, enforcement, knowledge-learning and policies. Within this cycle an autonomic/cognitive system is always aware of its environment/state (context), it makes decisions upon its operation and in accordance with established policies that specify goals/objectives and constraints on the actions to be taken, and finally it enforces them. In parallel, it exhibits the ability to learn from the obtained results and to build knowledge on the context and the used policies. This cycle can be seen as a variation of the well known “Monitoring Analyze Plan and Execute — Knowledge (MAPE-K)” model, i.e. the model of feedback loop that was proposed by IBM for promoting its vision of autonomic computing . This is also done in order to facilitate the identification of “hooks” that can be used to establish and evolve autonomic principles in the existing standards. Results of this analysis are summarized in Table 1.
ETSI RRS WG3 Overview — The ETSI TR102.682  defines a Functional Architecture (FA) for the Management and Control of Reconfigurable Radio Systems targeting at improving the spectrum and radio resources utilization. The following use cases are considered in this report. In the case of a terminal arriving in a new geographical place where it has no knowledge of the radio environment, the initial start-up use case describes how the network supports that terminal with information to efficiently find a suitable radio access. In the secondary spectrum usage use case, information is provided on secondary spectrum usage opportunities in a heterogeneous multi-RAT environment. The spectrum on demand use case focuses on flexible spectrum usage between operators where each operator is a primary user of a certain spectrum and operators can transfer certain parts of the spectrum for a period of time to another operator. For the radio resource optimization, users can be handed over from one RAT to another. In certain cases, such an optimization may need a terminal reconfiguration which can consist of a software download and/or a modification of the operating RAT for SDR-capable terminals. In order to enhance the overall performance of the network, base stations can be reconfigured as described in the network reconfiguration use case. The ETSI RRS FA with the specified functional blocks and their interfaces is depicted in Fig. 2. These blocks act in whole or in part in both network (NET) and terminal (TE) side. In general, the Dynamic Spectrum Management (DSM) provides the mid- and long-term management of the spectrum (e.g. in the order of hours, days) for the different radio systems. The Dynamic Self-Organizing Network Planning and Management (DSONPM) provides the medium and long term decisions upon
IEEE Network • November/December 2011
ETSI RRS FA Single- or multi-operator, heterogeneous context/(Self-)management distributed among network and terminal
Analyze — Plan
Decision making/ Optimization
DSM (medium- & long-term spectrum occupancy evaluations, provision of spectrum to RATs, reassignment and sharing/trading of spectrum, etc.) DSONPM (medium- & longterm QoS assignment to applications, traffic distribution to RATs, RAT activation, spectrum selection & radio parameters configuration etc.) JRRM-NET (short-term radio access-selection, handover, QoS/bandwidth allocation/ admission control)
CCM-NET (+underlying RATs)
DSM (spectrum related measurements) DSONPM (elements’ status and capabilities) JRRM-NET (QoS, cell load, radio conditions etc. through JJ-TN)
IEEE 1900.4 and 1900.4a Single- or multi-operator, heterogeneous context/(Self-)management distributed among network and terminal
3GPP SON immediate, product driven 3GPP centric self-organization
JRRM-TE (measurements, neighborhood & access selection info through JJ-TN)
RMC (RAN context information) WSM (regulatory context information and information from white space database) CBSMC (CBS context information)
TMC (terminal context information)
Information exchange among eNodeBs (Radio Link Failure reports, radio resource usage, load indicators, capacity indicators, access and access delay probability estimation, interference indicators) SH_MON_F (measurements) SH_MMF (Self-healing information, actions and results)
HO-triggering measurement reports, Access attempts
JRRM-TE (idle state access selection)
NRM (decisions for network reconfiguration and policies for terminals) CBSRM (decisions for CBS reconfiguration and policies for terminals)
TRM (decisions for terminal reconfiguration within framework of NRM and CBSRM policies)
OA&M and/or eNodeB (No SON-specific entities for Self-Configuring and self-Optimizing, SH_DG_F for Self-Healing)
No SON intelligence in terminal
CCM-TE (+underlying RATs)
RRC (RAN reconfiguration) CBSRC (CBS reconfiguration)
TRC (terminal reconfiguration)
QA&M (No SON-specific entities for Self-Configuring and Self-Optimizing, SH_TG for Self-Healing)
No SON intelligence in terminal
Mentioned, but out of the scope of standardization activities
Spectrum usage policies RAT selection policies NO strategies (load balancing, energy, etc.)
None (left to implementation of particular algorithm)
Spectrum assignment policies NRM radio resource selection policies CBSRM radio resource selection policies
None (left to implementation of particular algorithm)
Operator QoS requirements Network management policies Initial planned RACH configuration policies
Table 1. Summary of AWNM standards analysis.
reconfiguration actions of certain network segments. The Joint Radio Resource Management (JRRM) provides neighborhood information for the efficient discovery of available radio accesses and selects the best radio access for a given user. Finally the Configuration Control Module (CCM) is responsible for the execution of the reconfiguration of a terminal or a base station, following the directives provided by the JRRM or the DSONPM.
Analysis Context Awareness — The monitored context information varies between the different building blocks with respect to their purpose and the targeted timescale of their applica-
IEEE Network • November/December 2011
tion. Measurements reflecting current spectrum assignments, availability of spectrum bands, e.g. for trading etc., are specified as part of the DSM. The DSONPM defines a context acquisition function for monitoring the status and capabilities of the elements of the network segment they manage and the status of their environment. In a shorter time scale, information like the availability of access networks and their condition (e.g. cell capacity, load), requested QoS parameters (e.g. bandwidth, maximum delays), radio conditions (aggregated signal strength/quality) and user preferences are monitored by the JRRM. Additionally, the JJ-TN interface between JRRM-NET and JRRM-TE
Operator 2 SS
JR RAT 2
CR RAT 1
CR RAT 2
JR RAT 1
CR RAT 1
DSM - Dynamic spectrum management DSONPM - Dynamic, self-organizing planning and management JRRM - Joint radio resources management CCM - Configuration control module
Figure 2. ETSI RRS functional architecture (FA) for the management and control of reconfigurable radio systems.
also has the role to carry Neighborhood Information (e.g. on RAT, cell location and size, capabilities or their dynamic data) and Access Selection Information (policies and/or decisions). Decision Making — The DSM is responsible for spectrum occupancy evaluations, provision of available amount of spectrum to RATs, detection of long-term available spectrum bands for reassignment and sharing/trading of spectrum, derivation of economical parameters for spectrum trading, etc. The DSONPM decides on reconfiguration actions at the application layer (e.g. guaranteed QoS assignment to applications), network layer (e.g. traffic distribution to specific transceivers and corresponding RATs), and also lower/PHY layer (e.g. radio parameters configuration per RAT). The fundamental objective of the JRRM block is to make shorter term decisions upon: • Selecting the best radio access (initial access selection decisions made in the terminal side and handover decisions made in the network side) for a terminal, based on the context and policies described below. • QoS/bandwidth allocation/admission control, which are considered per user session or connection based on the requested user QoS applications. A subset of decisions can also be made in the terminal side (JRRM-TE) for the radio access selection, either wholly when in idle state or in coordination with the network, when in connected state. Enforcement — The CCM building block is responsible for the execution of reconfigurations of a terminal or a base station following the decisions made by JRRM or DSONPM.
Policies — Policies comprise an important asset of RRS FA. A first type provides the regulatory framework for the spectrum usage. The second type consists of policies provided from the network to the terminals for assisting the radio access selection and handover decisions. Further policies, denoted as NO strategies, guide the management decisions made by the DSONPM. For instance, a NO might choose to apply a load balancing strategy among its RATs, or an energy aware strategy including greener decisions, e.g. by minimizing the number of used transceivers and consumed power. Other policies designate rules to be followed in context handling. Knowledge-Learning — “Knowledge acquisition based on learning functionality” is described as an essential feature required for increasing reliability in management decisions and for addressing complexity and scalability issues. Implicitly, there is a clear tendency toward relying on machine learning for enhancing optimization and decision making (that cannot, of course, be part of standardization).
IEEE 1900.4 and 1900.4a Overview — The IEEE standard 1900.4  was published in February 2009. It considers three use cases: distributed radio resource usage optimization, dynamic spectrum assignment, and dynamic spectrum sharing. In the distributed radio resource usage optimization, one or several operators operate several RANs using the same or different RATs. Frequency bands assigned to these RANs are fixed. Also, reconfiguration of radio equipment on network side is not considered. In dynamic spectrum assignment, one or several operators operate several RANs using the same or different RATs in different frequency bands. To improve radio resource usage,
IEEE Network • November/December 2011
WS RAN TRC
Terminal Another TRM
CBS Another CBSRM Packet based network
1900.4a system RAN - Radio access network WS RAN - White space RAN CBS - Cognitive base station OSM - Operator spectrum manager NRM - Network reconfiguration manager TRM - Terminal reconfiguration manager CBSRM - CBS reconfiguration manager
RMC - RAN measurement collector TMC - Terminal measurement collector CBSMC - CBS measurement collector RRC - RAN reconfiguration controller TRC - Terminal reconfiguration controller CBSRC - CBS reconfiguration controller WSM - White space manager
Figure 3. 1900.4 and 1900.4a system architecture.
configuration of these frequency bands can be dynamically changed. In dynamic spectrum sharing, several RANs using the same or different RATs can share the same frequency band. IEEE standard 1900.4a  (expected to be published in August 2011) amends 1900.4 by defining additional components so as to enable mobile wireless access service in White Space (WS) frequency bands without any limitation on used radio interface. It concentrates on the dynamic spectrum sharing use case of the 1900.4 standard. Figure 3 shows the 1900.4 and 1900.4a system architecture. Eight entities are defined on the network side. The Operator Spectrum Manager (OSM) enables the operator to control Network Reconfiguration Manager (NRM) dynamic spectrum assignment decisions. The RAN Measurement Collector (RMC) collects RAN context information and provides it to NRM. RMC may be implemented in a distributed manner.
IEEE Network • November/December 2011
The NRM manages network and terminals for distributed optimization of spectrum usage. The RAN Reconfiguration Controller (RRC) controls reconfiguration of RANs based on requests from NRM, and it may also be implemented in a distributed manner. The White Space Manager (WSM) enables collaboration between a 1900.4a system and a 1900.4 system, provides regulatory context information to Cognitive Base Station Reconfiguration Manager (CBSRM), and enables communication between CBSRM and the WS database. The CBSMC collects Cognitive Base Station (CBS) context information and provides it to CBSRM. The CBSRM manages CBS and terminals for network-terminal distributed optimization of spectrum usage. The CBS Reconfiguration Controller (CBSRC) controls reconfiguration of CBS based on requests from CBSRM. Moreover, three entities are defined on the terminal side: The Terminal Measurement Collector (TMC) collects termi-
S o n
S o n
S o n
S o n
S o n
S o n
Itf-X2 S o n
S o n
Itf-X2 S o n
S o n
Itf-X2 S S o + o n n
S o n
Figure 4. High-level overview of SON architectural options.
nal context information and provides it to the Terminal Reconfiguration Manager (TRM). The TRM is the entity that manages its terminal for network-terminal distributed optimization of spectrum usage. This management is performed within the framework defined by the NRM radio resource selection policies and the CBSRM radio resource selection policies. Finally, the Terminal Reconfiguration Controller (TRC) controls reconfiguration of the terminal based on requests from the TRM.
Policies — The policy-based management approach is used in the 1900.4 and 1900.4a standards to implement distributed decision making. Three types of policies are defined: • Spectrum assignment policies generated by OSM and targeted to NRMs. • NRM radio resource selection policies generated by NRM and targeted to TRMs. • CBSRM radio resource selection policies generated by CBSRM and targeted to TRMs.
Analysis Context Awareness — 1900.4 and 1900.4a define a set of entities on the network side and the terminal side that are responsible for collecting required context information in a distributed and autonomous manner. In the network side RMC collects RAN context information, WSM collects regulatory context information and exchanged context information with the WS database, and CBSMC collects CBS context information. In the terminal side, TMC collects terminal context information.
Knowledge-Learning — There is neither implicit nor explicit reference to any knowledge/learning related requirements in the 1900.4 and 1900.4a standards. It is intentionally left to implementation of the particular algorithms so as not to limit the scope of the standard.
Decision Making — 1900.4 and 1900.4a define a set of entities on the network and terminal sides responsible for decision making. Each of these entities is autonomic to some extent in making its decisions. In the network side NRM is responsible for making reconfiguration decisions for heterogeneous wireless network components and for generating NRM radio resource selection policies, CBSRM for making CBS reconfiguration decisions and for generating CBSRM radio resource selection policies. In the terminal side TRM makes autonomic terminal reconfiguration decisions within the framework defined by NRM and CBSRM radio resource selection policies. Enforcement — After the decisions are made, some of them may require reconfiguration on the network side and/or terminal side. 1900.4 and 1900.4a define several entities on the network and terminal side that are responsible for controlling such reconfiguration. In the network side, RRC is responsible for controlling reconfiguration of RANs within a heterogeneous wireless network according to NRM decisions, and CBSRC controls CBS reconfiguration according to CBSRM decisions. In the terminal side, TRC controls terminal reconfiguration according to TRM decisions.
3GPP SON Overview — Whereas IEEE 1900.4/4a and ETSI RRS cover a quite similar heterogeneous management framework, 3GPP SON is closer to short-term implementations that cover a single, cellular RAT. Being introduced as part of the 3GPP LTE, SON provides an autonomic management framework as a key driver for reducing traditional high operational and capital expenditures during the entire network lifecycle, i.e. from planning and deployment to optimization and maintenance phases. Important SON functions are self-configuration, selfoptimization and self-healing (often complemented by functions such as self-diagnosis and self-protection). Self-configuration is responsible for the configuration of newly deployed nodes (eNodeBs) through automatic installation procedures for getting the necessary basic configuration for system operation. Self-optimization aims at adapting the network to variations in traffic, propagation conditions and modifications in the operating conditions such as the introduction of a new service. Self-healing intends to automatically detect and localize most of the failures and applies self-healing mechanisms to solve several failure classes. In general, SON in 3GPP is covered by RAN2, RAN3 and System Architecture 5 (SA5) groups, which focus on signaling between the eNodeB and terminal, on interfaces and protocols and on telecom management aspects, respectively. Most importantly, the 3GPP 36.902 release 9 recommendation  provides descriptions of nine use cases and solutions with regards to self-configuring and self-optimizing networks.
IEEE Network • November/December 2011
Specifically, they consist of Coverage and Capacity Optimization, Energy Savings, Interference Reduction, Automated Configuration of Physical Cell Identity, Mobility Robustness Optimization, Mobility Load Balancing Optimization, RACH Optimization, Automatic Neighbor Relation Function and Inter-cell Interference Coordination. The 3GPP 32.541 release 10 recommendation  describes Self-Healing concepts and requirements of Operations, Administration and Maintenance (OA&M) and identifies three main use cases, namely Self-Recovery of Network Element Software, Self-Healing of board faults and Self-Healing of Cell Outage. Ten logical Self-Healing functional blocks are defined in . Selectively, the Self-healing Input Monitoring Function (SH_MON_F) is responsible for monitoring the triggering events, such as alarms, and gathering information. The Selfhealing Diagnosis Function (SH_DG_F) does the appropriate analysis, and then the Triggering Recovery Action/s Function (SH_TG_F) triggers appropriate recovery actions to automatically solve the fault. The Self-healing Evaluating Function (SH_EV_F) evaluates the healing result and decides the next step accordingly. Finally, the Self-healing Monitoring and Management Function (SH_MMF) controls the whole process and provides the operator with necessary information. From an architectural perspective, SON algorithms may reside either in OA&M or in eNodeB. Communication between eNodeBs is allowed through the X2 interface, and between eNodeB and OA&M through the Itf-N interface. It is expected that a self-configuration subsystem will be created in OA&M to be responsible for the self-configuration of eNodeBs. For self-optimization and self-healing functions, they can be located in OA&M (centralized SON) or eNodeB (distributed SON) or both of them (semi-distributed SON). Figure 4 depicts the location/execution of centralized, distributed and semi-distributed SON functionality in a simplified view of the LTE network management structure. Different colors for OA&M represent vendor-specific OA&M systems, where standardizing interfaces between OA&Ms can guarantee inter-vendor operability. OA&M entities belonging to different domains communicate through Itf-P2P, namely the interface used between peer Domain Managers.
Analysis Context Awareness — 3GPP SON use cases and solutions define a set of measurements and performance data both in the network and terminal sides. In the network side, the following context awareness procedures are defined in several use cases. In Mobility Robustness Optimization, radio link failure (RLF) reports are exchanged between neighboring eNodeBs, such as identifiers of the cell in which the RLF occurred, of the cell where RL re-establishment attempt is made, and of the UE in the cell where RLF occurred. In Mobility Load balancing optimization, the eNodeB should monitor the load in the controlled cell and exchange related information with neighboring node(s) over X2 (or S1), such as the current radio resource usage and the current load using load and capacity indicators. In RACH Optimization, RACH load monitoring and other eNodeB measurements are performed. RACH optimization functions need to estimate access probability (AP) or access delay probability (ADP), in order to set RACH parameters, e.g. configuration and transmission power control parameters. In Inter-Cell Interference Coordination, reporting signaling and periods of uplink/downlink power and interference indicators are required for the coordination of eNodeBs in terms of preferences/priorities in resources. Furthermore, in SH_MON_F, measurements and monitoring of the defined Trigger Conditions are per-
IEEE Network • November/December 2011
formed, and in SH_MMF information is provided to the operator, including self-healing actions and results. In the terminal side, context awareness procedures exist either in Mobility Robustness Optimization or in RACH Optimization. In Mobility Robustness Optimization, HO-triggering measurement report messages (MRM) are sent by the UE. In RACH Optimization, UEs should report necessary information needed to estimate AP and ADP, such as the number of attempts needed to obtain access. Moreover, the UE indicates whether it has been subject to contention during the random access procedure. Decision Making & Enforcement — Depending on the function, decisions are made and enforced, e.g. for initialization of radio parameters, new site/service addition (self-configuration), or for updating radio (resource management) parameters (selfoptimization). No SON-specific functional entities are defined for decision making and execution of self-configuration and self-optimization. In self-healing, SH_DG_F that analyses/diagnoses and identifies the appropriate recovery actions and SH_TG_F that triggers the execution of recovery actions are defined. It can be also stated that there is no SON intelligence in the terminal side; it is exclusively hosted by the network. Policies — 3GPP SON considers the interaction between selfconfiguring/optimizing networks and OA&M. Through OA&M, policies are defined based on the operator QoS requirements (all SON use cases), the network management policies (Mobility robustness optimization), or the initial planned RACH configuration (RACH Optimization). Knowledge-Learning — There is neither implicit nor explicit reference to any knowledge/learning related requirements in the 3GPP SON standards.
Discussion and Challenges End-to-End Vision/Integration The described standards have provided elaborate specifications for functions, functional blocks and interfaces related to AWNM. However, they are confined to their own specific domain and this is in contrast with one of the main requirements toward the FI, which is to get rid of the inflexible, “in silos” type of management and ensure an end-to-end management vision. This can only be accomplished by extending standards to leverage openness and interoperability. What we have so far is a rather mature standardization activity regarding autonomic/cognitive aspects in the wireless access segment, but a lot to be done so that autonomic management systems can be federated across networks and management domains, thus providing an end-to-end management view. Two directions of integration/expansion are foreseen in order to achieve this end-to-end vision: toward the wired/core network segments and toward the “micro” wireless world including Internet of Things (IoT), as well. Multiple research efforts have focused on designing autonomic management architectures toward integrated wireless access and wired/core segments. Nevertheless, with the exception of the ETSI AFI Industry Specification Group (ISG), the objective of which is to develop pre-standard specifications for Autonomic network engineering for the self-managing FI (AFI), the standardization of autonomics in other targeted groups is actually in its infancy, e.g. the Network Management Research Group (NMRG) of the Internet Research Task Force (IRTF) which is a major contributor to the Internet Engineering Task Force (IETF) and definitely leaves room for enhancements w.r.t. a new management plane for the FI;
High level operator policies/goals
IETF SMI/ MIBs
DEN-ng TMF SID IETF PAWS
Extend policy frameworks
Knowledge and information (semantic interoperability)
ETSi RRS WG3
3GPP MTC ETSI M2M ISG
Extension/integration with “micro” wireless world, IoT Device-to-device 3GPP SON (thing-to-thing) ETSI TR 102.684
Autonomic (cognitive) wireless network management standards
ETSI AFI ISG Extension/integration with core/wired IEEE segments DySPAN 1900.4/4a
Wireless access segments
NMRG. (IETF) Wired/core segments
ITU-T SG13 FG-FN
Figure 5. Challenges and way forward.
or the ITU-T SG13 “Focus Group on Future Networks (FGFN)” which aims to collect, review and identify visions and design objectives for future networks. In the other direction, autonomic/cognitive management features should be expanded to cover a different type of communication, i.e. among devices (and/or things), as elaborated within Machine Type Communications (MTC) in 3GPP or Machine-to-Machine (M2M) ISG in ETSI. It must also be noted that certain interfaces in ETSI RRS (JJ-TT from JRRM-TE to JRRM-TE) and 1900.4/4a (TRM to TRM) might need to be expanded and further elaborated to achieve this goal. Recently, a new working item entitled “Feasibility Study on Control Channels for Cognitive Radio Systems” and numbered as TR 102.684 was also formed within ETSI RRS WG3 for extending the RRS FA toward this direction .
Information and Knowledge Aspects Considering the above described integration only from the communication point of view will not be sufficient. It is also necessary that this integration will happen by populating knowledge databases and sharing knowledge on what each autonomic node perceives from its surrounding environment. Although partially addressed as at least a requirement, e.g. in ETSI RRS, the main information exchange covers simple context data necessary for triggering and driving the decision making of the cognitive management entities. There is also a certain need for specifying models (structure) of the knowledge bases and interfaces for the collection, dissemination and use of this structured knowledge. Related to that and 1900.4a, IETF has recently initiated a standardization effort called Protocol to Access White Space (PAWS) databases so as to investigate the specification of a messaging interface between WS devices and the WS database. It is also very important to specify interfaces for sharing knowledge among autonomic/cognitive managers or otherwise, knowledge will not be reusable. At this point, the notion of information models is brought into the foreground. 1900.4 defined its own information (including policy) model. 3GPP is traditionally based on the notion of Integration Reference Points (IRPs) and designate, e.g. Common Object Request Broker Architecture (CORBA, Simple Network Management Protocol (SNMP) or Extensible Markup Language (XML) as
solution sets for implementing these IRPs. ETSI RRS did not specify nor adopt any model. Obviously this list spans larger and wider management domains including, for example, Structure of Management Information (SMI) based Management Information Bases (MIBs) of IETF, Common Information Model (CIM) of Distributed Management Task Force (DMTF), Shared Information/Data model (SID) of TeleManagement Forum (TM Forum), Directory Enabled Networks-next generation (DEN-ng), and many more proprietary and/or research oriented models. Since one of the necessary challenges in the FI is to integrate/federate the management systems of disparate network domains, semantic interoperability among their also disparate information models is a sine qua non. Toward this direction, the role of ontologies and semantic-related standards like World Wide Web Consortium (W3C) and/or the Organization for the Advancement of Structured Information Standards (OASIS) needs to be considered as well.
The Role of Policies When designing the future autonomic/cognitive networks, the role of operator control should not be underestimated. The inspection and control of the behaviors of autonomic elements need to be guaranteed by means of new techniques and functionalities that must gain the human trust. Policies are very important here. The human operator should have the means to set business level (technology agnostic) goals in terms of high level policies and let them be enforced in their self-x capable network by means of low level policies and configuration commands (technology specific). Automation of such a procedure is both desirable and challenging. The described standards comprise examples of “operator governed,” autonomic/cognitive management. They specify a framework for balancing among fully autonomic decisions and policy-based centralized control; however, they target a single application area or domain. In an integrated network world the realm of policies must also span multiple heterogeneous segments (e.g. wireless and wired) and associated technologies, which are properly managed by multiple autonomic/cognitive control loops. NRMs, DSONPMs, etc. and other types of autonomic managers will probably need to coexist and coordinate. Their already specified policy frameworks need to expand so as to define explicit goals and guarantee their con-
IEEE Network • November/December 2011
flict-free, stable and orchestrated behavior. The above discussion is summarized in Fig. 5.
Conclusions It is commonly recognizable that the envisaged benefits of FI will also be accompanied by equally remarkable complexity. This requires contemplating network management at the beginning of designing FI, so as to avoid phenomena of reduced manageability that exists in the current Internet. Autonomics and autonomic network management appear as the most viable solution for tackling such complexity. In particular, considering wireless networking and its evolution, there has been an excessive growth of research in AWNM. However, business impact and quick adoption can only be achieved by standards that will ensure openness and interoperability. To this effect, this article presents the advances of major bodies in the standardization of AWNM and at the same time it discusses the challenges that need to be addressed so that AWNM will efficiently complement the self-managing FI.
Acknowledgment The work of the authors at the University of Piraeus has been performed in the context of the OneFIT project (Opportunistic networks and Cognitive Management Systems for Efficient Application Provision in the Future InterneT (www.ict-onefit.eu)) which is supported by the European Community’s Seventh Framework Program (FP7). This article reflects only the authors’ views, and the Community is not liable for any use that may be made of the information contained therein.
References  N. Samaan and A. Karmouch, “Towards Autonomic Network Management: An Analysis of Current and Future Research Directions,” IEEE Commun. Surveys & Tutorials, vol. 11, no. 3, 2009, 3rd Quarter 2009, pp. 22–36.  R. Boutaba et al., “Recent Advances in Autonomic Communications,” Guest Editorial, IEEE JSAC, vol. 28, no. 1, Jan. 2010, pp. 1–3.  S. Schuetz et al., “Autonomic and Decentralized Management of Wireless Access Networks,” IEEE Trans. Network and Service Management, vol. 4, no. 2, Sept. 2007, pp. 96–106.  M. Mueck et al., “ETSI Reconfigurable Radio Systems — Status and Future Directions on Software Defined Radio and Cognitive Radio Standards,” IEEE Commun. Mag., vol. 48, no. 9, Sept. 2010, pp. 78–86.  IEEE Std 1900.4™-2009, IEEE Standard for Architectural Building Blocks Enabling Network-Device Distributed Decision Making for Optimized Radio Resource Usage in Heterogeneous Wireless Access Networks, Jan. 2009.  IEEE Standard 1900.4a-2011 for Architectural Building Blocks Enabling Network-Device Distributed Decision Making for Optimized Radio Resource Usage In Heterogeneous Wireless Access Networks — Amendment: Architecture and Interfaces for Dynamic Spectrum Access Networks in White Space Frequency Bands, 2011.  3GPP, “Evolved Universal Terrestrial Radio Access Network (EUTRAN); Selfconfiguring and self-Optimizing Network (SON) Use Cases and Solutions,” 3rd Generation Partnership Project (3GPP), TR 36.902, June 2010, available: http://www.3gpp.org/ftp/ Specs/html-info/36902.htm.  ETSI, TC RRS, “Reconfigurable Radio Systems (RRS); Functional Architecture (FA) for the Management and the Control of Reconfigurable Radio Systems,” Technical Report 102 682 v1.1.1, July 2009.  3GPP TS 32.541 V10.0.0 (2011-03), Technical Specification, 3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Telecommunications Management; Self-Organizing Networks (SON); Self-healing Concepts and Requirements (Release 10).  J. O. Kephart and D. Chess, “The Vision of Autonomic Computing,” IEEE Computer, vol. 36, no. 1, Jan. 2003, pp. 41–50.  FP7/ICT project OneFIT (Opportunistic networks and Cognitive Management Systems for Efficient Application Provision in the Future InterneT) (ICT-2009257385), July 2010–Dec 2012, Website: www.ict-onefit.eu, Apr. 2011.
IEEE Network • November/December 2011
Biographies KOSTAS TSAGKARIS [M] ([email protected]
) received his diploma and his Ph.D. (Ericsson’s awards of excellence in Telecommunications) from the School of Electrical and Computer Engineering in National Technical University of Athens, Greece. Currently, he works as a senior research engineer and adjunct lecturer at the Department of Digital Systems in University of Piraeus, Greece. He has been involved in many research projects and published more that 100 papers in the areas of design, management and optimization of wireless autonomic/cognitive networks. He has also contributed to EU and US standardization committees and working groups such as RRS and AFI-ISG in ETSI and IEEE DySPAN/P1900.4, where he has also served as Technical Editor of the published standard. STANISLAV FILIN [SM] ([email protected]
) works as an expert researcher in NICT since 2007. He is chair of IEEE 1900.7 WG on WS Radio. He is voting member of IEEE DySPAN-SC, IEEE 1900.1 WG, IEEE 802.11 WG, IEEE 802.15 WG, and IEEE 802.19 WG. He served as technical editor in IEEE 1900.4 WG. He served as NICT delegate in ETSI TC RRS and ITU-R WP 1B and WP 5A. In 2009 he received IEEE SA SB award for contribution to the development of IEEE standard 1900.4-2009 and NICT award for contribution to IEEE 1900.4 standardization. JENS GEBERT ([email protected]
) received his Diploma in Electrical Engineering in 1993 from the University of Stuttgart. In the same year, he joined Alcatel as a software engineer for Mobile Systems. He currently works for Alcatel-Lucent in Bell Labs Germany in the Wireless Access Domain on Cognitive Radio Systems. He is strongly involved in the architecture work within different EU FP7 projects as well as in ETSI TC RRS standardization activities, in which he leads the Working Group 3 (WG3). M ARKUS M UECK [M] ([email protected]
) received the diploma degrees of the University of Stuttgart, Germany and the Ecole Nationale Supérieure des Télécommunications (ENST), Paris, France in 1999, the Doctorate Degree of ENST Paris in 2006 and became Adjunct Associate Professor of Macquarie University, Sydney, Australia in 2009. He was a Senior Staff member and Technical Manager at Motorola Labs, Paris, France from 1999 to 2008. Then, he was Standardization Manager at Infineon Technologies, Germany and now has identical roles at Intel Mobile Communications, Germany since 2011. He is chairman of ETSI RRS. PANAGIOTIS VLACHEAS ([email protected]
) received the Diploma and Ph.D. degree from the National Technical University of Athens, School of Electrical and Computer Engineering, in 2003 and 2010 respectively. Currently, he is a senior researcher at the Telecommunication Networks and Integrated Services laboratory of Digital Systems Department of the University of Piraeus. His research interests are in the area of Wireless networks, Radio Resource Management, Admission Control, Packet Scheduling, Reconfigurable, Cognitive and Self-Organizing networks, Optimization Theory. GEORGE ATHANASIOU ([email protected]
) received the diploma in Computer and Communications Engineering from University of Thessaly in 2005. He obtained his M.Sc. and Ph.D. degrees in Computer and Communications Engineering from the same University, in 2007 and 2010 respectively. Currently, he is a senior researcher at the laboratory of Telecommunication Networks and Services of the Digital Systems Department of the University of Piraeus. His research interests are in the design and performance evaluation of wireless and fixed broadband networks. VERA-ALEXANDRA STAVROULAKI ([email protected]
) is an Assistant Professor at the Department of Digital Systems of the University of Piraeus. She received a diploma in Informatics from Athens University of Economics and Business in 2000, and a Ph.D. degree, in Electrical and Computer Engineering, from the National Technical University of Athens in 2004. She has 10 years of experience in European and national research and development projects. Her main interests include cognitive management functionality for autonomous, reconfigurable user devices (operating in heterogeneous wireless networks), service governance and virtualization. HIROSHI HARADA ([email protected]
) is director of the Smart Wireless Laboratory at NICT. He joined the Communications Research Laboratory, Ministry of Posts and Communications, in 1995 (currently NICT). Since 1995, he has researched software defined radio (SDR), cognitive radio, dynamic spectrum access network, smart utility network, and broadband wireless access systems on the microwave and millimeter wave bands. He serves currently on the board of directors of the Wireless Innovation (formerly SDR) Forum and as chair of IEEE Dyspan Standards Committee (formerly SCC41 or IEEE P1900) and vice chair of IEEE P1900.4 and IEEE P802.15.4g and TIA TR-51. He was chair of the IEICE Technical Committee on Software Radio (TCSR), 2005-2007, Japan.