New Network Management Software will Improve the Quality of IPTV Service
A group of researchers from Portuguese higher education institutions, Carnegie Mellon University, and Portugal Telecom are concluding the development of a novel network operation and management framework, which aims to improve the delivery of Internet Protocol Television (IP/TV) content to customers. The deployment experiments with data from Portugal Telecom (PT) are having positive results and shortly will move to real time data.
This is a result of the Next Generation Network Operations and Management (NeTS) project, carried out in the scope of the Carnegie Mellon Portugal Program, funded by the Fundação para a Ciência e a Tecnologia. The project follows a new approach to the operation and management of IP/TV networks and combines the expertise of a cross-disciplinary team led by Ricardo Morla, the principal investigator of the project in Portugal who is with the Faculdade de Engenharia of the Universidade do Porto and INESC TEC, and Hyong Kim, his counterpart at Carnegie Mellon University.
It is well known that networks are difficult to manage and operate. Companies are spending more on the daily management and operations of their networks than investing on developing and launching new IT services. According with the principal investigators of this project, studies have shown that scheduled maintenance and upgrades can account for more than 30% of network outages in Tier 1 ISPs, and operator errors are common and can be the root causes for more than 50% of failures in computer systems and networks. The excessive cost of network management is mainly due to the large scale of networks, the heterogeneity of different technologies, and their subtle interactions.
The NeTS project is based on hierarchical network abstraction modeling, structure learning of probabilistic graphical models, and kernel-based learning technologies. To automatically try to understand the nature of the data that goes through the line to a set-top box TV, the researchers gathered “insights from network abstraction, statistical and heuristic machine learning methods intended to build an intelligent network abstraction model,” explained Ricardo Morla.
The team is trying to diagnose the problems that occur in the network broadband access lines, modeling and classifying the data in order to improve the quality of the service. The records say that the connection is the major problem, specifically driven by connectivity loss. “We try to classify the different types of problems underlying that connectivity loss, which could be power failure, noise, and electromagnetic interference in the lines, for instance,” explained Ricardo Morla.
To model the time pattern of the events the research team is extracting important information from the available network management data. Concerning these specific topics, two papers have been submitted. The first one is on improving metrics to diagnose anomalies in the Digital Subscriber Line (DSL) access network using Renyi entropy. The second paper is on learning how to automatically classify different types of DSL anomalies. In both papers the main concern was to help improve the quality of the DSL service provided by the operators. Another paper on this topic was also submitted to the journal IEEE Transactions on Network and Service Management, which addresses the application of Support Vector Machine (SVM) classifiers from a class to this problem.
The Project Team Involves Academia and Portugal Telecom
In addition to the approach and the use of different models to track the behavior of the networks, an innovative aspect of this project is its team, composed by researchers that have knowledge of networking and data.
The team involves researchers from the Instituto de Telecomunicações (IT), INESC TEC, Faculdade de Engenharia of the Universidade do Porto (FEUP), Faculdade de Ciências of the Universidade do Porto (FCUP), Instituto Superior Técnico of the Universidade Técnica de Lisboa (IST/UTL) and Carnegie Mellon University (CMU). The Portuguese telecommunications company Portugal Telecom (PT) also collaborates in this project, providing real data to classify the synchronization loss. “The tests we made with the data provided by PT reached results as good as 99 percent of reliability, which is very good”, said the Portuguese principal investigator. It is expected that PT could be using this software in the future.
Apart from the principal investigators, the project also involved two post doctoral researchers, Simon Malinowski and Angelos Marnerides, and a dual degree Ph.D. student on Electrical and Computer Engineering, Chen Wang. Her work is on the quality of the video provided to users, namely in “trying to optimize how the video servers are instantiated so that the quality of the service that we provide to the users improves”, said Ricardo Morla. The student is co-supervised by the two PIs of this project. The CMU PI is looking into the configuration of the network and into cloud-based approaches.
The software, that runs automatically, reads the data from the log files with management events, processes them and stores the information on a database. “We built a webpage that provides information on the data in a graphical way,” added the researcher. The way data is structured in the program facilitates stakeholders conclusions and decisions, hopefully allowing the operator to proactively correct problems before the customer could even notice them.
Concerning the future, Ricardo Morla has already something in mind to continue his research on the same topic when the NeTS project is concluded. “We will definitely put into place new proposals based on this project's achievements. We are planning to propose a project on management of cloud and video infrastructure,” he said.