Autonomic network management approaches have not been widely adopted, mainly due to significant unsolved challenges. Challenges include technical complexity, lack of consistent models and knowledge bases describing the system, and the difficulty of evolving management methods and processes. Autonomic approaches often operate a closed control loop. Such loops enable dynamicity and are often intent driven, where system goals and requirements are declared, then automatically accomplished and maintained. These loops continuously monitor and analyze large amounts of information to infer knowledge about the system. Representing the knowledge as semantic graphs is well suited to automated inference, enabling hidden relationships, strategies and understanding to be identified. When applied in an autonomic network management system this automatic discovery of additional knowledge can be used in several ways to inform and improve intent driven closed control loops. This paper describes the design and evaluation of an ontology to represent and help interpret, validate and apply high level goals or 'intents' as part of a closed control loop. This approach enables these intents to be enforced, satisfied and maintained. The ontology forms part of a framework which generates graph-based data from network monitoring information collected in a commonly used network/cloud monitoring service (Prometheus). The ontology also models intents relative to the monitoring knowledge. Furthermore, the model has the capabilities to allow the monitored network to adapt, then helps plan how to continuously satisfy and maintain the intent. Finally, the ontology and framework are applied in a real-life use case, which relates to Quality of Service (QoS) assurance for a 5G Telecoms Network Slice. The use case is designed to motivate and demonstrate the usefulness of the approach.

Today we see more and more systems support intent languages, frameworks and interfaces. The incorporation of these approaches provide elements of abstraction between the request and the execution. However mapping is still required to get from the abstract to the real. This mapping often introduces elements of rigidity into the system. This work presents a flexible approach to intent driven systems through an interpreter for intent realization. The mechanism of the interpreter is demonstrated through interactions with mock slice manager functions running on an open source, functions as a service platform. The interpreter accepts intent messages containing a request in the form of an English sentence, along with required parameters for the request. This information is then used to identify the appropriate functions available to the interpreter. These functions are described in the Functionality Template model. Once identified, the Functionality Template guides the building of an action in response to the intent message.

The linked data generation process is a complex process which involves multiple stakeholders in the definition of mapping artefacts which transform source data into linked data representation. The creation of these mapping artefacts is error prone, and the quality of the artefacts should be assessed prior to the generation of the linked data. Producing high quality mappings will result in high quality linked data and provide higher confidence to linked data consumers. Furthermore, the source data of these mappings should be regularly monitored to detect data changes which could impact the quality of the resulting linked data dataset. A process designed to offer fresh and high-quality data will benefit both data consumers and producers. Applying the process to a real-world use case demonstrates the feasibility and effectiveness of the process. In this paper we describe a process designed to improve quality within the linked data generation process. Furthermore, we describe the application of the process to a real-world cloud monitoring use case

This demonstration focuses on the comparison and identification of conflicts between independent goals described through intent.Our approach, implemented in the Adaptive Policy EXecution system, is presented in 3 stages: 1) Intent generation, describing the structure of an intent 2) Intent comparability, detailing the requirements allowing for the effective comparison of goals and 3) Conflict resolution, detailing the process of identifying appropriate responses in accordance with already established intent goals. This demo will showcase each of the stages highlighting both the positive aspects and limitations of the approach.

Services such as interactive video and real time gaming are ubiquitous on modern networks. The approaching realisation of 5G as well as the virtualisation and scalability of network functions made possible by technologies such as NFV and Kubernetes pushes the frontiers of what applications can do and how they can be deployed. However, managing such intangible services is a real challenge for network management systems. Adaptive Policy is an approach that can be applied to govern such services in an intent-based manner.In this work, we are exploring if the manner in which such services are deployed, virtualized, and scaled can be guided using real time context aware decision making. We are investigating how to apply Adaptive Policy to the problem of optimizing interactive video streaming delivery in a virtualized environment. We utilise components of our previously established test bed framework and implement a single layer neural network through Adaptive Policy, in which weights assigned to network metrics are continuously adjusted through supervised test cycles, resulting in weights in proportion to their associated impact on our video stream quality. We present the initial test results from our Perceptron inspired policy-based approach to video quality optimisation through weighted network resource evaluation.

Configuration management of large scale networks is a complicated business. This is especially the case for Converged Network Operators (CNOPs). Here, networks are a mix of fixed and wireless access, MAN and backbone networks, as well as interchanges with other operators.The Recursive InterNetworking Architecture (RINA) offers a number of abstractions and claims that their application will simplify not only network operations but also their management. In the H2020 project ARCFIRE we have taken these abstraction, built a Distributed Management System (DMS), developed a configuration management strategy, and tested all aspects against a set of important Key Performance Indicators (KPIs). We examined the performance of node and network creation, validation of a configuration, complexity of the developed software, and capabilities to automate configuration management tasks. This paper provides an introduction to our work, including a discussion of our (aggregated) test results.

Taking an autonomic approach to management and using closed control loops has been the subject of much research in the Network management community since the early 2000s. It is fair to say that Network Management system developers and users have not adopted Autonomic Management approaches very widely. Most network management systems continue to use an ITU TMN inspired layered approach to management. In recent years, a trend towards implementing autonomic management and closed control loops on management systems built using a TMN architecture has emerged in practice. This trend is requirement driven; an autonomic approach is taken when there is no other option for implementing a feature. It is clear to see a closed control loop approach being taken to implement C-SON (Centralized Self Organizing Networks) features in 4G network management systems in the early 2010s. Autonomic approaches are even more apparent in systems such as ONAP that implement SDN and NFV orchestration. However, the implementation of closed control loops is often pragmatic and rigid, focused on the feature being delivered. Providing systemized support for control loops is in its infancy and has much to learn from the extensive autonomic management literature This paper surveys the current state of autonomic management in practice and outlines some research challenges that must be addressed to allow it to be systematically supported in current management systems, with a particular focus on ONAP.

This demo focuses on demonstrating features of the Adaptive Policy Execution (APEX) system. APEX is a carrier- grade, production ready, scalable policy engine implementing, based on published theory, universal and immutable policy infrastructure. The demo will showcase the main APEX features, from authoring via deployment to runtime; with three demo use cases. All software and features are available on Github.

Since the 1970's it has been acknowledged that a complex system can be broken into (a) its invariant functional parts (mechanism), and (b) the externalized choices for how the system should behave (policy). Policy-based management's main objective is to separate and externalize the decisions required by a system from the mechanisms provided by the system, and provide a way to define and evaluate these decisions. A few decades later, we have today a plethora of different policy models and even more policy languages - plus tooling - offering policy-based solutions for virtually any use case and scenario. However, policy-based management as a standalone domain has never been evaluated in terms of which parts are variant / invariant, i.e. which parts of policy-based management can be domain-, model-, language-, usecase-independent. In this paper, we introduce and define a formal universal policy model that does exactly that. The result is a model that can be used to design, implement, and deploy immutable policy infrastructure (engine and executor) being able to execute (virtually) any policy model.

Due to the increase in the dynamicity, programmability, scope and complexity of modern networks there is a greatly increased requirement that network management systems control, orchestrate and manage networks in a much more automated and dynamic manner. This drive towards automation and dynamicity requires autonomic network management that continuously analyses network state and continually steers the network in accordance with changing high level goals and policies. As dynamicity increases, it is proving increasingly difficult to test and validate the analytics routines and policies that drive today's network management systems. With more automation, the potential for unanticipated network incidents increase, for example where multiple automation features interact and conflict. There is no substitute for seeing how a network management feature actually performs in a real network, ideally allowing iterative authoring/validation development cycles. However, due to the high stakes involved in degrading or disrupting network performance, this is not usually feasible until the very final testing and deployment stages. The next best option is a testbed that accurately represents a live network scenario to support authoring and validation development cycles in a low-risk environment. In this work we present our experiences of building a networking testbed that incorporates an emulated network, a production- grade network controller, an analytics function, and a policy execution environment. This allows users to develop policies for adaptive (closed loop) management of a realistic emulated network. We also present two scenarios where the testbed is used to emulate and mitigate against a temporary and prolonged failure occurring on a network.

Video in all its forms is probably the most important service carried on networks today and few would argue that video quality assurance is one of the most daunting network management challenges. Quite often, video optimization strategies and their decisions are an integral part of either the video protocol (e.g., dynamic adaptation of rate and quality) or the distribution systems (e.g., multi-level caching architectures). A unified method of assuring video services is a formidable task, especially as the world prepares for the adoption of 5G network concepts and the associated complexity. In parallel, policy has been proposed as an approach for managing domains in a flexible and adaptive manner. In this work, we describe our approach to use adaptive policy to externalize the goals and decision making of optimization strategies in the form of a network resource evaluation and path selection experiment aimed at video service quality assurance. In this paper, we present our approach, outline our initial implementation and discuss our preliminary results.

5G networks will be the first real converged networks supporting a plethora of different services, each with their own requirements. A static best-effort approach is no longer sufficient. Extreme flexibility and dynamicity is required, yet costs must be drastically reduced. The only way that these conflicting goals can be achieved is with vastly increased automation in the provision and operation of our future 5G networks. In this paper we briefly discuss the facilitators, goals and challenges for 5G networks. We identify some of the places where automation is not just helpful, but is in fact required for 5G to become a reality. We go on to present a conceptual approach for modeling and achieving autonomic operations and management in 5G networks positioning modern policy-based management as a key enabler for autonomic 5G network management.

The COMPA (Control, Orchestration, Management, Policy, and Analytics) adaptive control loop realizes an automation pattern that can operate recursively at many layers in a carrier network. An overall COMPA autonomic control loop can orchestrate functions, themselves implemented as COMPA autonomic loops. Thus the COMPA automation patter can recurse right down to resource level in a network. One of the most exciting application areas for the COMPA automation pattern is in assuring mobile network security. The recursive nature of the pattern is the ideal mechanism for automating monitoring and root cause analysis of security threats to networks. In this paper we present a Proof of Concept of a COMPA compliant system for a Distributed Denial of Service (DDoS) scenario. The system monitors, performs root cause analysis, and mitigates a DDoS attack. The system was built by integrating a number of existing components that were deployed as VNFs. Our experiences of using the system were that the system could handle a DDoS attack quickly and automatically. In addition, the system was very flexible to build and deploy.

The formal structure of information models and the controlled manner of accessing and changing such models brings both flexibility and control when managing network elements. However, keeping information models synchronized and consistent across network elements and management systems is a challenging task. Today this problem is exasperated with the advent of ephemeral network functions and elements and also by the need for distributed scalable cooperating management functions running in containerized cloud deployments.

For some time now the volume of data traffic in Mobile Telecommunications Networks has far outweighed voice traffic and most users are more concerned about the quality of their data connection than their voice calls. Adequate traffic service requires analysis of traffic flows to identify network issues affecting users' services. This can be achieved by placing network analytics probes throughout the network, but it would be better to have a small number of probes at just key locations in the Mobile Core Network. We show how to use core network traffic analysis to identify which user owns and which base station hosts each traffic flow, particularly where users and their connected traffic flows may dynamically switch between base stations or radio technologies.

The advent of “Soft Networking”, where networks are composed of virtual nodes and links, promises to dramatically ease the definition and deployment of networks whilst allowing network applications that are limited only by the imagination of the developers of those applications. In such a dynamic environment, the Autonomic Management pattern supervised by policies has been recognised as holding more promise for management of Soft Networks than traditional techniques. We have proposed Dynamic Adaptive Policies as an approach to give classic policies the dynamicity and flexibility to manage such networks and whatever applications are running on them. In this paper, we describe our ongoing work on Apex, an engine that executes and administers Dynamic Adaptive Policies in a scalable and distributed manner.

Management systems increasingly consume events, giving them a more real-time view of the networks they are managing. Providing an event forwarding mechanism between the management system component that consumes events and application instances is a challenge. Such a mechanism must forward events in a manner that ensures correct delivery and is scalable, reliable, and efficient. This paper describes an approach that uses pools of event consumers and application instances to receive events being sent to a network management system from network elements in a telecommunication network. The use of pools allows the task of processing incoming events to be balanced across the members of the pool. The size of the pools can be automatically modified to cope with the current event load. The approach has been used in a proof of concept management event processing system installed on two live mobile networks.

New technologies are changing the world of communication networks and even more so their management. Cloud computing and predictive analytics have removed the need for specialized compute hardware and created products that continuously search for and find insights in management data. Virtualization of networks and network functions, SDN and NFV, are beginning to be mature enough for production networks resulting in much more flexible and dynamic networks. IoT and M2M traffic and new customer demands are driving new thinking and demands for 5G networks. Almost every aspect in the control and management of networks has seen new dimensions of flexibility and dynamicity, with the notable exception of the policies that drive them. This paper discusses the need to add adaptiveness to classic policies, describes a novel approach for adaptive policies, shows how adaptive policies will form part of future network frameworks and architectures, and finally discusses early use cases developed for mobile operators.

Monitoring the massive volume of data streaming from managed nodes in Telecommunication networks reacting in a timely manner is increasingly critical for modern Telecommunications Operations Support Systems (OSS). Given the large number and the varieties of the nodes in a telecoms network, the streaming monitoring data is naturally diverse and the volume is often at scales of multiple millions data points each second. These data are well modelled using formal syntaxes (e.g. Management Information Bases), making formal semantics and automated reasoning a viable solution for Telecom data modeling and correlation. This paper proposes an approach that will leverage recent developments in Semantic Reasoning and Big Data. The paper introduces how we propose to use RDF stream reasoning methods for real time event correlation, combined with MapReduce technologies in order to decentralize the large number of reasoning and correlation tasks that need to be undertaken in real time. The proposed approach is currently being implemented and will be evaluated using the diverse data types and volumes that are expected.