John Keeney

A recommender system architecture for predictive telecom network management

Current telecom networks generate massive amounts of monitoring data consisting of observations on network faults, configuration, accounting, performance, and security. Due to the ever increasing degree of complexity of networks, coupled with specific constraints (legal, regulatory, increasing scale of management in heterogeneous networks), the traditional reactive management approaches are increasingly stretched beyond their capabilities. A new network management paradigm is required that takes a preemptive rather than reactive approach to network management. This work presents the design and specification of E-Stream, a predictive recommendationbased solution to automated network management. The architecture of E-Stream illustrates the challenges of leveraging vast volumes of management data to identify preemptive corrective actions. Such design challenges are mitigated by the components of E-Stream, which together form a single functional system. The EStream approach starts by abstracting trace information to extract sequences of events relevant to interesting incidents in the network. After observing event sequences in incoming event streams, specific appropriate actions are selected, ranked, and recommended to preempt the predicted incidents.

Towards real-time management of virtualized telecommunication networks

The idea of virtualizing network functions is driven by recent advances in network-focused hardware. In 2012, several large telecommunication operators issued a call to action for Network Function Virtualization (NFV)1. The underlying idea is that every network service currently delivered on proprietary, application specific hardware should be deliverable using virtual machines. This means that routers, firewalls, load balancers and other network devices run virtualized on commodity hardware. Consequently, ETSI is extending this idea to mobile networks. Here, parts of the core and the radio access network will be virtualized. The immediate advantage is that any Virtual Network Function (VNF) can now be deployed, re-deployed and undeployed in the same way as any traditional virtual machine. Thus, NFV will result in more dynamic and agile networks than seen heretofore. However, this will raise a number of serious issues in managing these future networks. In this paper, we examine issues and challenges in orchestrating these virtualized functions and their interconnections to provide a more agile mobile telecommunication network.

E-stream: Towards pattern centric network incident discovery and corrective action recommendation in telecommunication networks

With the technological evolution in telecommunication networks, performance requirements such as better coverage, higher bandwidth, and lower latency have been pushed to new horizons. However, as a direct result network complexity has increased dramatically over the recent years, and with this complexity manageability has suffered. This paper presents the architecture of the E-Stream project which aims to support Next Generation Operations Support Systems. E-Stream applies dimension reduction, data mining, and recommender system techniques in order to handle very high volumes of management events, identify and predict network incidents, and recommend candidate corrective actions to domain experts in Network Operations Centres.

Dynamically adaptive policies for dynamically adaptive telecommunications 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.

Context-aware heterogeneous network performance analysis: Test-bed development

The rapid growth in mobile and heterogeneous network has led to a massive increase in data traffic adding to the already significant challenges in network performance. Operators already face the problem that thousands of network incidents (e.g. dropped calls or packet delivery failures) remain unresolved due to the technological limitations. It is recognized that the use of network emulation/simulation test-beds are an valid approach to validate the network performance prior to rollout. This paper introduces the development of a context-aware heterogeneous network test-bed which allows network administrators to flexibly control both network topology and network traffic, as well as analyze network performance through output network events. The developed test-bed includes both open-source simulators and low-cost real life equipment. In order to demonstrate the benefit of the test-bed, cell congestion and handover use cases are implemented. SIP services are delivered over various radio network technologies including LTE, UMTS, WiMAX, GSM and WiFi. Various types of network events are provided as output for potential users, i.e. operators, network administrators, etc.

An algorithm for load balancing in network management applications managing Virtualised Network Functions

In network management applications, load balancing has typically been achieved by manually configuring the application in-situ or using configuration information generated using offline tools. As networks increase in scale and heterogeneity, management applications are being designed to take advantage of scalable computer hardware and often have multiple instances, each of which bears a portion of the management load of the network elements or functions being managed. The communication between management applications and the network elements they are managing is often stateful, posing a challenge for load balancing because connections must be managed if responsibility for managing some network elements change from one instance to another during load balancing. In this paper, we describe an algorithm that enables load balancing of telecommunication management applications for an arbitrary number of VNFs (Virtual Network Functions). The algorithm, triggered by NFV (Network Function Virtualisaton) orchestration, is horizontally scalable and there is no single point of failure once two or more management application instances are running.

A framework to leverage domain expertise to support novice users in the visual exploration of Home Area Networks

Advances in modern technologies have afforded end-users increased convenience in performing everyday activities. However, even seemingly trivial issues can cause great annoyance for the ordinary user who lacks domain expertise of the often complex systems that underpin these advances. A key challenge lies in assisting non-expert users to express their requirements of an obscure and complex system. This research proposes a semantic approach by using domain expert knowledge to enable real time semantic up-lift in supporting novice end-users to understand and control the complex dynamic systems they must manage. This presents a significant opportunity to increase user satisfaction and reduce associated support costs. This semantic approach has been designed and implemented in an early prototype of our Home Area Network Monitoring System (HANMS). This paper presents a detailed description of the current state of the research, an initial evaluation, and future work.

i-MagNet: A real-time intelligent framework for finding specific needles from needle stacks

Currently the volume of telecom network management data is expanding exponentially, mainly due to the explosive growth in the number of communicating devices along with the increase in heterogeneity of the networks. Such scale of data obsoletes the traditional approach of extracting offline analytics from the network traces governed by some pre-defined schemes. In order to increase the efficiency of the Operations Support System (OSS) and gain in-depth understanding of the generic relationship between network entities, the monitoring data needs to undergo large-scale deep analytics processing. In this paper we present i-MagNet, an integrated analytics framework developed with the popular real-time stream processing paradigm Storm. The components of i-MagNet intelligently micro-batch segments of incoming streams to enable high-throughput online analytics of management trace streams. Inter-dependence metrics (temporal and statistical) are exploited to extract contiguous event subsequences, which can then be independently examined as part of a network incident analysis system.

Apex: An engine for dynamic adaptive policy execution

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.

Qualitative comparison of open-source SDN controllers

Telecommunication companies with expensive networks may become the biggest beneficiaries of SDN; however, in contrast to traditional routers, the development of SDN controllers is driven by open-source projects with involvement of the industry. Two prevalent projects in SDN development are the OpenDaylight and the ONOS controllers. These SDN controllers are advanced in their development - having gone through a number of releases - and have been described as being useful for a large number of use-cases. In this work, we compare and evaluate these controllers, in particular their northbound interfaces, by configuring them for a representative use-case, port-mirroring.