Traditionally, network management tasks manually performed by system administrators include monitoring alarms based on collected statistics across many heterogeneous systems, correlating these alarms to identify potential problems or changes to management policies and responding by performing system re-configurations to ensure optimal performance of network services. System administrators have a narrow focus of factors impacting network service provisioning and performance due to the heterogeneity and scale of generated underlying network events. However, self-adaption principles are conceptual approaches for autonomously managing such complex distributed systems. Network management systems that harness such principles can dynamically and autonomously optimise the operation of network services, responding quickly to changes in user requirements and underlying network conditions. In this paper, we present a novel self-adaptive network management framework that takes advantage of a recursive network architecture for a simpler and more comprehensive application of ontologies, semantic web rules and machine learning to automatically adjust network configuration parameters to provide more optimal QoS management of network services. We demonstrate the applicability of the approach using a content distribution network (CDN) operating over such a recursive network architecture.

Virtualization is an enabling technology that improves scalability, reliability, and flexibility. Virtualized networking is tackled by emulating or paravirtualizing network interface cards. This approach, however, leads to complexities (implementation and management) and has to conform to some limitations imposed by the Ethernet standard. RINA turns the current approach to virtualized networking on its head: instead of emulating networks to perform inter-process communication on a single processing system, it sees networking as an extension to local inter-process communication. In this article, we show how RINA can leverage a paravirtualization approach to achieve a more manageable solution for virtualized networking. We also present experimental results performed on IRATI, the reference open source implementation of RINA, which shows the potential performance that can be achieved by deploying our solution.

In Service-Oriented Architecture (SOA), dedicated intermediate nodes called load balancers are usually deployed in data centers (DC) in order to balance the load among multiple instances of an application service and to optimize the resource utilization. However, the addition of these nodes increases the installation and operational cost of DCs. These load balancers distribute incoming flows to multiple outgoing ports usually by hashing them. Several techniques are used in order to select the outgoing ports e.g. round robin, queue length, feedback from neighbors etc. Such load balancing approaches do not considergetting live feedback from the service end and therefore are not able to dynamically change the amount of allocated resources. In this paper, a distributed load management scheme isproposed for service oriented networks based on the currentInternet architecture. In this scheme, lightweight interconnected management agents are used to decide the availability for a particular service instance and help in optimal distribution of the flows. The proposed scheme can also be applied in other emerging internetworking architectures such as RINA.

Software Defined Networks (SDN) has taken the world by storm. Only a few years old as technology, most of the big players have SDN in their product portfolio or in their strategic roadmap. SDN has changed the way we virtualize the network fabric in data centers, provided new features for cloud computing, and arguably plays a big role in facilitating Network Function Virtualization (NFV). Looking ahead, SDN has the potential to make the network disappear altogether, similar to Mark Weiser’s vision for computing. However, while SDN’s main contribution is a new south-bound interface for TCP/IP flow control, little work has been done on the north-bound interface for the interaction with and the management of an SDN network. Two essential items are still missing. First, SDN does currently not provide means to expose network capabilities to applications (e.g. a QoS cube), thus it fails to bridge the gap between the network and services. Second, SDN does not help to advance network management while introducing many new challenges for it. In this paper, we start discussing the disappearing network and discuss how to address the two missing items in the progress. Our contribution is to examin the Recursive InterNetworking Architecture (RINA) as an evolutionary step for SDN, which we present that in the form of three use cases.

Resource access control policies for an organization are often derived from best practice standards or from high level business policies. To ensure that access control is enforced effectively, these business policies need to be translated into deployable system configurations or lower level policies for multiple diverse systems. These target policy representations require experts to coordinate and collaborate so that business policies are fully supported. It is difficult and cumbersome to effectively ensure that all access control policies are enforced with the desired effect and in a consistent way, particularly given that there may be many people editing policies and that business policies can change over time. We present a language driven approach that abstracts access control policies into a clear and structured set of rules defined using terms familiar to a non-systems expert, which may then be realized into multiple levels of abstraction. Our proof of concept system uses Language-Driven Development (LDD) techniques to transform high level business policies into device specific policies that can be enforced by multiple access control system types. Our scenario examines the application of access control to instant messaging communications and network server access, two systems with different access control configuration languages.

We present a dominance detection algorithm as part of a policy authoring process that makes extensive use of semantic models to perform a novel dominance detection of access control policies, where groups of deployed policies are considered in unison to discover redundancy. The approach is targeted towards the pre-deployment stage of the policy authoring process and aims to help prevent the introduction of redundant policies into the system. To achieve this, semantic queries are executed over instances of new and deployed policy elements in order to select matching elements for further analysis. The semantic queries may return a large number of deployed policy elements so we present an algorithm that prunes the search space to reduce the problem size. We show that for large sets of deployed policies, we can discover relatively large sets that are considered dominant.

Federations are viewed as persistent agreements between organisations that enable them to share information or capabilities in a controlled manner. Policy based management techniques can be used to support federations of service providers in two ways: 1) policies alleviate the need for expensive human attention in the federation set-up and maintenance processes, and 2) policies can operate and maintain the federation in a more automated, but guided fashion requiring less manual intervention by system administrators. This thesis presents a federation policy authoring process that allows for specification and consistency analysis of federation policies that adhere to a federation model during refinement. Federation policies are refined into multiple lower-level device language implementations using model-driven language development techniques. During refinement, and as part of the policy authoring process, the consistency analysis process uses ontologies and semantic web rules to retrieve deployed policies for consistency analysis using a policy element match algorithm. The element match algorithm analyses groups of related policies to detect relationships between them; this contrasts with state-of-the-art pairwise policy analysis which is not capable of detecting all inconsistency cases. In addition, an approach to tailor policy evaluation for enterprise social networks is proposed to cater for different policy execution environments ranging from low to high risk security environments; this approach is shown to increase evaluation performance.

policy conflict analysis process for the analysis of newly specified federation-level policies against previ­ ously deployed 10caVfederation policies. The process is generic in the sense that it can be employed by each domain participating in a federation to help maintain the consistency of their local system with that of the federation. The conflict analysis process utilises an information model and associated ontology for representing both the static and dynamic application-spec ific aspects of the local operating environment and the federation to aid in the detection of potential inconsistencies. It employs two algorithms, one for the selection of previously deployed policies related to the federation-level policy and the other for analysing the returned policies against the federation-level policy. The selection algorithm reduces the number of deployed policies required to be retrieved for analysis against any newly specified federation-level policy, while the conflict analysis algorithm detects inconsistencies relating to the conditional element of a policy rule. We discuss a

To support rapidly evolving business models, communications network management systems are increasingly being federated to provide more flexible, end-to-end service management. In the future, such federation will need to be achieved dynamically, thus management systems will need to incorporate capabilities supporting negotiation of federations and management of their lifecycle. We discuss how federations can be governed via negotiated federation-level policies, that should be consistent with the relevant local policies of individual federation members. We describe a policy authoring process, outlining the steps to be taken when local or federation-level policies are created, modified or withdrawn. As this process depends on the presence of a rich system model for policy analysis we describe an extension to DEN-ng that models governance of federated domains. Finally, we outline a case study relating to inter-organisation XMPP federations to illustrate the policy authoring process.