Network Monitoring Applications
Based on IoT System
Andrej Kos, Urban Sedlar, Janez Sterle, Mojca Volk,
Janez Bešter
Laboratory for Telecommunications
Faculty of Electrical Engineering, University of Ljubljana
Ljubljana, Slovenia andrej.kos@fe.uni-lj.si Abstract— We present applications for network monitoring based on intelligent communication platform that can also be used to support various usage scenarios related to future internet of things. Applications presented include real time DSL access line monitoring and IPTV monitoring, correlated with lightning reports. The solution is used in the field of proactive monitoring, enabling the operator's helpdesk and field technical teams to pinpoint the cause of service degradations.
Keywords— network monitoring; IoT; DSL; IPTV, lightning; visualisation; correlation

I.

INTRODUCTION

Today, telecommunications operators offering access
(i.e. xDSL, FTTx) and services (i.e. IPTV, Internet access,
VoIP) are confronted with the problem of SLA (Service Level
Agreement) and minimization of operational cost.
In xDSL (including FTTN and local loop shortening), the fulfillment of SLA is an issue due to crosstalk, which is the interference between two pairs of a cable. Increasing the number of subscribers on the same cable, shortening the last mile, increasing the speeds, the mixture of different broadband technologies and interference with wireless devices causes degradation of service already on physical layer. Operators are failing to provide services that they have sold to users in the past. As the need for bandwidth is becoming larger, operators are forced to move access devices closer to the users.
In IPTV, the fulfillment of SLA is an issue of all devices and systems that provide the service, starting with set-top-boxes, home installation, access device, access network, aggregation network, core network, service delivery platform and head end.
Therefore, the operators have to deploy efficient monitoring and correlation tools for all network layers to keep the current network operation under control and proactively detect changes and trends in their network. Data collected and correlated enables commercial sectors to better plan, define and offer new packages, technical departments to better plan and upgrade networks, and helpdesk teams to pinpoint the cause of eventual service degradations, thus being able to respond faster and more accurately.
Such a monitoring and alarm correlation tool is a combination of performance monitoring on different network layers and alarm monitoring with addition of an event

Marko Bajec
Laboratory for Data Technologies
Faculty of Computer and Information Science,
University of Ljubljana
Ljubljana, Slovenia marko.bajec@fri.uni-lj.si correlation and pattern matching expert tool. As the data are collected from the network and from other external sources, such as the environment, and there are huge amounts of data, such systems are also known as IoT (Internet of Things) or big data. Advantages of such approach are that (i) it enables 100 % penetration (monitoring the entire network) and (ii) monitoring the entire network line, including the CPEs, means real end-toend operator network coverage.
As the DSL solution is well described in [1], we put focus on IPTV in this article.
II.

IOT PLATFORM

The IoT platform is an open communication platform for data integration and development of data-driven and event-driven services to be used in various communication and service environments. The platform is designed universally and can be scaled in order to accommodate different scenarios (i.e. different fields of application), further enhancements and modifications. The majority of platform components are generic and can be used in various use cases (where simple user scenarios apply), while domain specific components are kept to minimum (required for complex user scenarios) [2, 3].
One of the key functionalities of the platform is the possibility of merging data from different sources, use cases and scenarios (Fig. 1) [4, 5]. This prospect paves the way towards enhancing data with data from other domains.
Consequently, new use cases and new data integrations are possible. The platform architecture enables simple and straightforward inclusion and integration of new domain scenarios. Based on that and through open interfaces, development of new and innovative services is easy and straightforward. Fig. 1. Merging data from different sources. It enables common storage, preprocessing and visualizations.

18th European Conference on Network and Optical Communications & 8th Conference on Optical Cabling and Infrastructure - NOC/OC&I 2013
ISBN: 978-1-4673-5822-4, July 10-12, 2013, Graz, Austria

69

The platform architecture (Fig. 2) comprises event-driven components (i.e. real time analysis, business activity monitoring, and alarming), data-driven components (i.e. persistence, interoperability, expert analysis and business intelligence), sources and external systems. The inputs to the platform are data and event sources; sensors and devices forwarding measured data into the platform, i.e. applications based on service and event driven architectures and legacy systems, and trusted external data sources, i.e. databases.
Events are placed in the event channel. From there they are dispatched to the event processing kernel and the database.
Event processing kernel provides for event filtering, identification of noticeable events, pattern matching, time series analysis, and event correlation. The outputs are the socalled generated or output events, representing important information for various recipients, i.e. alarms and indicators.
These events are placed into outbound queue to make sure all concerned recipients are informed about particular events. The events are placed into two queues for the use of internal and external subscribers. Processed events can be used internally or externally. Inside the platform, monitoring is possible with generic monitoring dashboard providing key performance indicators (KPI). A basic KPI is an individual event that a user wants to monitor, while complex KPI consists of more
(correlated) events. Users can select KPIs and define type and mode of visualization (graph type, refresh rate, labels, etc.).
Special type of KPI, which can be forwarded to an email, short message service (SMS), or start an application, is an alarm.
The processing kernel provides for: duplicate elimination, data enrichment, complex pattern matching, prediction,
External subscribers and web applications events via external output queue. These external applications, mobile devices, etc.

entity resolution, expert analysis, and forecasting. are notified about are, for example,

We use DSL system in combination with OpenMN as a sensor network. Each individual DSL line represents a sensor node that provides required information and sends it to IoT platform. More details about the solution can be found in [1].
Therefore, in this section we only present the most representative case.
In Fig. 3, the values of actual downstream and upstream speeds on a given port are compared to the required speed of the user profile. For each user, a speed is set in a user profile; also, the minimum and the maximum tolerance is defined.

Fig. 3. DSL line monitoring. Values of actual speed on a given port is compared to the required speed of the user profile within tolerance.

IV.

IPTV

The second subsystem, in which we monitor in real-time is the IPTV system. IPTV systems are substantially more complex classic broadcast systems. They consist of a chain of at least 6 elements to each user and the multitude of support systems. The architecture of the IPTV system is shown in
Fig. 4.

Fig. 4. IPTV architecture. Complete system is substantially more complex classic broadcast systems.

Fig. 2. IoT platform architecture. It consists of internal building blocks and external sources and systems.

III.

DSL

We designed and implemented a solution for real-time domain monitoring of access line and access network. It consists of multi-service access node (MSANs) and management system. MSAN (in our case Iskratel SI300) has optical, DSL, fixed wireless and POTS/ISDN types of access.

70

Because of the maturity and reliability of the traditional broadcasting technologies, the anticipated user experience relating to the use of IPTV services is very high. The users are especially intolerant of errors in video content and poor service accessibility. The causes for faults and a lowering of quality can appear anywhere in the chain of providing the IPTV service; Anything from faulty content at the source, faults in the operator’s backbone network, faults in the access network, faults in the users’ home network, to errors in the video decoding process or functioning of the terminal equipment. In addition, there is a complex and technologically heterogeneous communication infrastructure between the provider of the
IPTV service and the user. This means that the possibility of

NOC/OC&I 2013, ISBN: 978-1-4673-5822-4

error increases proportionally with the number of network building blocks and technologies, and the capacity of the network connection to the user.
For the aforementioned reasons, monitoring of the users’ quality of experience (QoE) is one of the key diagnostic mechanisms in managing an IPTV system and services, used for an overall control of the system vitality, monitoring the quality of guaranteed services and identification, localization and elimination of network and application faults.
Typical monitoring systems that allow for the supervision of vital indicators of the quality of an IPTV system and services Key Performance Indicator – KPI, Key Quality
Indicator – KQI, have many flaws and limitations.
Firstly, the QoE is a subjective and individual measure of satisfaction of the end user, which demands time consuming empirical measurements of guaranteeing IPTV services, conducted on a sample set of end users, or an automated modeling and a QoE grade with objective methods of measuring the qualitative network and application parameters.
Secondly, the automated objective systems for measuring the vitality of the IPTV system available today are typically based on obtaining relevant data from a limited number of systematically chosen nodes in the network, which enables the monitoring of technical parameters of the quality of service, but it does not ensure the adequately sampled service context for a realistic evaluation of the application QoE. The collection of comprehensive data about the functioning of a system and the state of the topologically distributed termination points is not a trivial issue and has not been satisfyingly solved.
Our solution consists of a system for monitoring the network and the IPTV services, with advanced techniques for contextualization of quality and reliability of the IPTV services’ performance at the application level, with the use of techniques and principles of IoT. We designed a distributed system for the advanced measurement of KPI and KQI of an
IPTV network and services, as well as their in-depth analysis for the needs of assessing relevant contextualized events that have an effect on the quality and reliability of IPTV services.
The designed system encompasses the following architectural segments: (1) A software agent in an IPTV STB with the possibility of measuring application and network KPI/KQI; (2)
An IoT platform for gathering, storing and analyzing data, as well as a visual display on the system control panel; (3) A centralized system for agent management (control of the software agent on STB, setup of measuring functionality, measurements, logging and sending data to the platform).
Modeling and evaluation of QoE in IPTV systems are possible, based on measuring low-level network parameters
(QoS), but this approach does not offer accurate results due to the nondeterministic video flow and time-conditioned intra-event relevance, predictive and two-way predictive frames. More accurate information about the quality of an
IPTV user experience can be obtained directly from video flow decoding function of the IPTV terminal device. This calls for the dissection of the decoding process, isolation of relevant parameters in real time, with the help of data flow mining

NOC/OC&I 2013, ISBN: 978-1-4673-5822-4

techniques and forwarding of the collected statistical values of the video flow to the central data storage.
To diagnostically localize fault sources, there is a need for correlation of the captured and enriched measurements, as well as the physical and logical topological map of the IPTV network. The segment of potentially interesting metrics and additional information contains measurements of network QoS, streaming and application STB KPI (amount and type of decoding video and audio error, IGMP group, switching times, amount and type of network error), metadata about network elements (e.g. geographical location, network addresses, interface logic), network tree topology diagram, etc.
We use techniques of statistical analysis on the aggregated data, which, with knowledge of the network topology, allow for fault localization over the entire provider-user chain. Such information is useful to the operator for monitoring of the network state in real time (interactive control panel with identification of the root cause of issues in the network) and for long-term statistical error analysis, based on geographical region, network hierarchical level or TV channel.
Enrichment of acquired data with data from the operator’s backend systems allows for further analysis according to type and manufacturer of network equipment, and integration with the helpdesk service system, where the time from the user report to the cause identification is drastically shortened.

Fig. 5. IPTV error graph. Percents of error seconds are shown.

Fig. 6. IPTV monitoring. Dots with the same size represent a local area router. The color of the dot represent the level of errors (in this case, all dots are green, which means the level is below minimal treshhold). The red dots show the location and the strength of lightnings.

71

V.

VII. CONCLUSION

LIGHTNING

We developed a solution to correlate and visualize lightning information with DSL data and parameters and IPTV data and parameters. The IoT platform fetches the data from the trusted external server that collects and, over an API, exposes the information about lightning. The API provides information about location, location confidence range and time of the lightning [7,8]. Combining this information allows us to visualize the situation and correlate the events.

We presented network and service monitoring based on a generic light weight, open-source based real-time IoT platform.
It is a combination of a performance monitoring on different network layers and alarm monitoring, with addition of an event correlation and pattern matching.
With the solution, the operators get efficient monitoring and correlation tools for all network layers to keep the current network operation under control and proactively detect changes and trends in their network. Data collected and correlated enables commercial sectors to better plan, define and offer new packages, technical departments to better plan and upgrade networks, and helpdesk teams to pinpoint the cause of eventual service degradations, thus being able to respond faster and more accurately.
As IoT based applications are only starting to rise on all fields of life and business: from optimizing production processes, network monitoring, event correlation to a health and well-being.

Fig. 7. Example od lightning alerts. Similar table is used for any kind of alarm. VI.

RESULTS

Based on IoT approach and the real-time information from different sources, including last end-user device in a chain, we get and overview of whole network performance (CPEs, home, access, aggregation, core, service, inter-domain). Numerous graphs can be shown and alarms can be triggered, based on different combinations of KPIs, from statistical data to root cause discovery [6].
As an example, aggregated results from DSL, IPTV and lightning are shown in Fig. 8. Currently, in Telekom
Slovenije’s network, we have approx. 100.000 agents (sensors) enabled, with 7.000 agents sending data. Red dots show the location of the lightning, with the size of the dot representing the strengths. Green dots represent router locations. Depending on the number of errors, the green dots change color in real, so we have also the visual representation of IPTV error rate.
Graphs on the right show DSL performance.

The future brings us more user-friendly language to define
KPIs, more cross-domain data fusion, event correlations, visualizations and stream-mining.
In the long run, we anticipate future internet of things to develop into the so-called collective adaptive systems that will consist of many units/nodes, with individual properties, objectives and actions. Decision-making will become more distributed and the interaction between units will lead to an increase in unexpected phenomena. The operating principles will go beyond existing monitoring, control and correlation tasks, taking into account the diversity of objectives within the system, and the need to reason in the presence of partial, noisy, out-of-date and inaccurate information.
Many additional scenarios and applications will grow. Data collected has already become big-data.
ACKNOWLEDGMENT
The authors would like to thank company Telekom
Slovenije for cooperation on the research and development project "Automated system for triple-play QoE measurements", company Iskratel and the Slovenian Research Agency for co-financing the project “Quality of service and quality of experience measurement and control system in multimedia communications environments”, Ministry of Education,
Science, Culture and Sport and the Open Communication
Platform Competence Center.
REFERENCES
[1]

Fig. 8. Correlation of DSL, IPTV and lightnings. Based on different sources, different graphs and tables are shown. KPIs that are a correlation of other
KPIs can be monitored. Actions can be started based on the KPIs

72

[2]

Kos A., Pristov D., Sedlar U., Sterle J., Volk M., Vidonja T., Bajec M.,
Bokal D., and Bešter J.: Open and Scalable IoT Platform and Its
Applications for Real Time Access Line Monitoring and Alarm
Correlation. In: S. Andreev et al. (Eds.): NEW2AN/ruSMART 2012,
Lecture Notes in Computer Science 7469, pp. 27–38; Springer,
Heidelberg (2012)
Kos, A., Sedlar, U., Peternel, K., Volk, M., Sterle, J., Zebec, L., Vidonja,
T., and Bešter, J.; Odprta komunikacijska platformaIoT. VITEL - 25.

NOC/OC&I 2013, ISBN: 978-1-4673-5822-4

[3]
[4]

[5]

delavnica o telekomunikacijah, "Internet stvari", 12. 5. - 13. 5. 2011, pp.
1–5 (in slovene language)
Inteligoo Platform: www.inteligoo.com
Šubelj, L., Jelenc, D., Zupančič, E., Lavbič, D., Trček, D., Krisper, M., and Bajec, M.: Merging Data Sources based on Semantics, Contexts and
Trust. IPSI BGD Trans. Internet Res.; Vol. 7, No. 1, pp. 18–30 (2011)
Lavbič, D.. Rapid ontology development model based on rule management approach in business applications. Informatica (Ljublj.),
Mar. 2012, vol. 36, no. 1, pp 115-116

NOC/OC&I 2013, ISBN: 978-1-4673-5822-4

[6]

[7]
[8]

Sedlar, U., Volk, M., Sterle, J., Sernec, R., and Kos, A.: Contextualized
Monitoring and Root Cause Discovery in IPTV Systems Using Data
Visualization. IEEE Netw. Mag., Special Issue - Computer Network
Visualization; 1–9 (2012)
SCALAR System Network: http://www.scalar.si/en
Kosmač, J., Djurica, V., and Babuder, M.: Automatic Fault Localization
Based on Lighting Information. In: Power Engineering Society General
Meeting, IEEE: 8-22

73

This page is intentionally left blank.

74