Self-Awareness of Autonomic Systems

Position Statements

• Trust Establishment and Assessment Mechanisms in Autonomic Environments
• Engineering control levels in an Autonomic Network
• Self-management for cooperative transmission
• Self-management in service deployment by autonomic systems
• Security, as a built-in feature in Autonomic Systems
• Autonomic Management in the Future Internet
  
• Issues of Future Autonomic Networks
• Increasing Context Awareness in Autonomic Networks

TEAM in AE: Trust Establishment and Assessment Mechanisms in Autonomic Environments

 Institute of Communications and Computer Systems (ICCS) [Full text as PDF]

• " Irrespective of the specific scope of a network type (autonomic, wireless, core network, etc.) several mechanisms and operations require that nodes establish some type of communication among them,either for realizing collaboration or for implementing fundamental operations."
• "Malicious nodes may exploit lack of cooperation to intelligently attack legitimate nodes individually, or they may exploit cooperation capabilities in order to fake identities and create various types of exploitations."
• "it is necessary to establish a framework under which legitimate nodes of a network will be able to negotiate and trust each other. According to the designated negotiation and trust mechanisms, legitimate behavior will be possible to be evaluated and permitted, while malicious operation will be realized and prevented." via role concept, social-like environments, evolution of the trustworthiness,

• "A coherent,mathematical, self-contained framework for trust is required to quantify all the aforementioned and allow the accurate design and analysis of trusted systems " -dynamic iterative trust vector is proposed based on Markov Random Field method.
• "Until now the interdependencies between the context and trust have not been sufficiently addressed, but as various threats are related to context awareness, the context information (anonymised f personal) whenever available can offer opportunities to establish and manage the trust relationships more efficiently"
• "In contrast to trustworthiness, the reputation of a node is always defined among a set of nodes. It equally reflects the subjective views of a set of nodes."
• Trust assessment = "quickly respond to the question of whether the correct trust values have been assigned to the correct nodes."

Questions: A) How paradigms and solutions already applied in social networking, either web based social networks like facebook, goodread or application based P2P overlay networks, can be utilized into a pure autonomic environment? B) Could trust models already developed by social science be infused into autonomic networking bringing suchnetworks and infrastructures closer to the end user? C) Could user interfaces techniques regarding human machine interaction developed and largely adopted by web users be utilized by autonomic networking in order to make these systems more trusted by the end users?

Engineering control levels in an Autonomic Network

tmit.bme.hu (Full text as PDF)

• Autonomic Networking is an emerging paradigm to make this happen. The idea is to organize much of the network functionality into a hierarchy of so called control loops: the system constantly monitors for misbehavior or a change in the operational environment, then, after thorough analysis, a management decision is made in order to respond to the shift in network state based on aggregated system information, and finally the adjustment of the system is performed accordingly
• a fundamental task to determine the optimal level at which control decisions should be made (different problems require control loops of different “extent”. On one hand, a certain amount of information is needed to make a decision, which pushed decision making upwards in the hierarchy. On the other hand, resource efficiency or timeliness require to make decision making as fast as possible, which requires quick, low level decisions)
• Trade-offs and areas of interest:
  • In case of addressing, the structure and distribution of the address space is a major issue.
  • The routing and searching tasks on large scale networks may call for unstructured addressing like in social networking [2].
  • Scheduling of different time scales (packet, connection, process etc.) on a decentralized basis is an open issue.
  • Trust-based data forwarding inspired by human acquaintance networks is an interesting future challenge [3].
  • High performance autodefense and security mechanisms can have strict timing constraints, which imply reduced control loops.
  • Finally, fault tolerance and self-healing with respect to system performance is an interesting point.

Self-management for cooperative transmission

Telcordia Poland (Full text as PDF)

• cooperative techniques were introduced where the direct transmission between the source and the destination node is additionally enhanced with the aid of an additional retransmission performed by the relay node
• However, in the case of multi-tier set-ups is seems rather difficult for one node to take care of the overall control. What is more there would be an additional control overhead to be handled which could affect the system performance. The idea behind the concept of self-management for cooperative transmission would be then to make it feasible for the different groups of nodes to employ some relevant self-management techniques. As a consequence each node would know which group to attach to, which resources to use and what transmission technique(s) to apply. For this reason the nodes would have to be aware of their context in the network and they would have to exchange all the relevant information.

Problem Statements of self-management in service deployment by autonomic systems

Velti S.A.,(Full text as PDF)

• Two domains to study: the network domain and the context domain
• The need to define and specify common components in traditional management architecture still remains ultimately valid for self-management in autonomic service deployment however the framework should be enhanced with new/updated performance indicators or metrics to feedback the latest service generation quality requirements
• underlying realms of Internet of Services and Internet of Things are now interacting through a common network/context self-managed while on a service dependent way.
• There is a need to define a semantic-based framework for the description, composition and orchestration of autonomic enabled services
• Context awareness in networks and services still remains a key prerequisite for the realization of autonomics communication vision. Its ultimate factor of success depends highly on the self-awareness capability that in the case of service awareness requires a re-consideration of existing framework in order to incorporate service /application relevant features and parameters as well

Security, as a built-in feature in Autonomic Systems

Fraunhofer FOKUS, MOTION (Full text as PDF)

• the solutions suggested to deal with the security issues, in particular cryptographic techniques and access control can not be applied directly to Networks of the Future and in particular to autonomic systems for the three reasons:
  • flexibility: the need to support heterogeneity in access technologies: (Mobile, wireless, Ethernet, DSL and Cable, in networking paradigms (ad hoc, Peer to Peer (P2P), sensor networks, mesh, etc.), in applications (composed multimedia e-commerce , e-business, software as a service,etc.), in providers (network, content and service)
  • context awareness:Self-protection in autonomic networks can be achieved only through self-reconfiguration, which, in turn, heavily relies on the data collected from monitoring (context-awareness). Being aware of service requirements and potential threats, the system can adjust itself in a timely manner according to the network policies. In this case, the system will be able to provide the right service at the right time to the right user, detects attacks and misuse and takes actions to protect the network.
  • autonomicity:Autonomic security is an advanced step comparing to “reconfigurable” security. The former means adding more automation in the entire reconfiguration process and make the security system self-responsive. Similarly to autonomic systems, this can be achieved only through a control loop that monitors, analyzes and reacts.

Autonomic Management in the Future Internet

Telecommunications Software & Systems Group and Fraunhofer FOKUS (Full text as PDF)

• further investigation is required in a number of supporting areas such as Information Modelling and Ontological Engineering [2].To assist the development of advanced autonomic control algorithms, further investigation is also required in the application of bio-inspired autonomic systems [4] to various scenarios such as traffic engineering.
• A vital aspect of any autonomic management system is the ability to monitor and gain awareness of the state of the network. To this end, further research is necessary in in the area of traffic monitoring and analysis to assist autonomic management of a variety of applications such as QoS aware admission control [5] and network planning [6]. The ability to reason over and understand traffic patterns can greatly improve network management from the point of view of anomaly detection, fault detection and diagnosis, intrusion detection, traffic classification and characterisation.
• Within this domain of research, the need to evaluate and extensively test proposed architectures and algorithms must be facilitated through the development of real testbeds (IMS, IPv6) and network simulations (OPNET).

Some Issues of Future Autonomic Networks

Beijing University of Posts and Telecommunications and Fraunhofer FOKUS (Full text as PDF)

• some problems can also be brought by the introduction of the self-* characteristics in the dynamic environment such as the convergence rate of the autonomic control-loop, especially in the heterogeneous and mobile environment, the lack of the learning ability of the policy-based control-loop, the security and cost-effectiveness of the intrinsic monitoring techniques in the current autonomic networking technologies
• At the aspect of control/decision plane, the current designs of control loop only provide primary autonomic functions: the policy-based autonomic control loop lacks for learning ability, which restricts the autonomic capability of the network. Optimized decision is often difficult to achieve when decision element make a decision according to end-to-end performance goal.
• Contexts provide essential information to autonomic network nodes to make decisions. Network contexts acquirement technologies have received a lot of attentions. However, user behavior model, contents of applications and service features are also important context information for autonomic nodes to optimize network performance and QoS/QoE of user.
• In order to build the monitoring infrastructures securely with high performance, the security of the underlying monitoring mechanism should be carefully studied and the costs of large-scale monitoring activities should be evaluated. Furthermore, the large scale monitoring activities will produce massive data that need further processing and aggregating into the global performance information for the autonomic networks and services
• in the heterogeneous wireless environments, the different characteristics of the heterogeneous access technologies, together with the time-varying feature of wireless channels, result in a vibration during the information collection for the control-loops. In other words, the convergence rate of the control-loop in the autonomic networks can be low.
• NEEDED RESEARCH 1: Introducing the cognitive technology into autonomic networking architectures: The research on the cognitive network architectures design for end-to-end goals, which target to be self-adaptive, self-management, self-optimization; cognitive routing research based on future networks which support mobility; QoS autonomic management of cognitive networks Cooperation and communications among cognitive nodes
• NEEDED RESEARCH 2: Generalized context aware technologies: Constitute a packet prioritization strategy which orders application packets based on their contributions to guarantee the application’s QoS. According to the capability of the network and the receiving device, choose the most suitable transmittal strategy. Adaptively choose the route and the appropriate responder with less cost to satisfy the demand of the user and the content of request.
• NEEDED RESEARCH 3: Autonomic mobility management with different granularity support Because mobility management in multiple wired/wireless environments must be a multi-criteria based problem, key factors awareness and analysis is the important basic. Therefore, following problems should be solved to enhance the mobility management in the autonomic networks: (i) Awareness of interface status for a multiple access mobile terminal. (ii) Key factors analysis on multiple access networks selection and handoff. Mechanisms, protocols and policies are needed to be designed in the future autonomic networks to support mobility with different granularity.
• NEEDED RESEARCH 4: Fast control-loop convergence method for mobility management; The changes of wireless channel is very fast, this results in that much attention must be paid to the convergence rate of the control-loop in the autonomic networks. Therefore mechanisms for increasing the convergence rate in the context of mobility management must be studied. In order to enhance the mobility management functions, especially to realize fast control-loop convergence, collaboration and interaction of multiple mobile nodes are needed. Hence, modelling and mechanisms for interaction, communication and collaboration among multiple nodes should be studied.

Increasing Context Awareness in Autonomic Networks

 Greek Research and Technology Network (GRNET) (Full text as PDF)

• Autonomic monitoring is a vital process in (autonomic) networks since it allows network components to assess their own state and the overall network conditions. However, predefining the monitoring scheme can be inefficient in heterogeneous environments, taking into consideration the constant changes in the network topology and the diversity of the interconnected systems. Therefore, the monitoring components within an autonomic network should be continuously adapted, in a flexible manner, to an ever changing network infrastructure (self-configuration). By increasing the context-awareness of monitoring data exchanged by autonomic nodes it is possible to efficiently sense network conditions and the level of provided services and proceed to corrective actions (self-healing). Also, context-aware information is easily used in order to take decisions according to the specified administrators’ policies (self-optimization).
• The incorporation of the autonomicity concept to communications systems makes necessary the development of extensible context models. These enable the efficient representation of available information, needed for handling and distributing it. Ontology engineering has been proposed as a formal mechanism for (i) reducing the complexity of managing the information needed in network management and autonomic systems and (ii) for increasing the portability of the services across homogeneous and heterogeneous networks. Learning and reasoning techniques have to be used to support intelligent interactions among the specified entities. Ontologies will function as core components for automated analysis of enterprise-wide event data, will be based on user-defined rules, trigger corrective actions for healing the system, deal with policy based goals on a higher abstraction level and provide new levels of functionality. These issues have to be studied in more depth as autonomic networks are coming into reality
• A generic monitoring framework based on autonomic principles has to be designed. It would be responsible for controlling monitoring protocols and mechanisms and enabling advanced node/network functions (such as QoS, mobility) to access context rich information necessary to realize complex control loops. Today, information and data models are not capable of representing the detailed semantics required to reason about behaviour. It is necessary to augment our use of knowledge extracted from information and data models with ontologies. The proposed information model and ontology have to be extensible.

autonomic-communication.org is hosted by Fraunhofer Fokus