Irena Atov

• Doctor of Engineering
• Architect at Microsoft

Large-scale video conferencing services incur significant network cost while serving surging global demands. Our work systematically explores the opportunity to offload a fraction of this traffic to the Internet, a cheaper routing option offered already by cloud providers, from WAN without drop in application performance. First, with a large-scale...

Future networks, empowered with Artificial Intelligence (AI) and Machine Learning (ML) capabilities, are expected to become cognitive, autonomously thinking, learning, remembering and adapting to ever changing conditions. Knowledge acquisition and intelligent decision making are enabled, effectively supporting next generation networks in achieving...

Welcome to the February issue of the Data Science and Artificial Intelligence for Communications series of IEEE Communications Magazine. The series continues to grow in popularity among academicians, researchers and practitioners who apply artificial intelligence (AI) and machine learning (ML) techniques to solving problems across diverse domains o...

The articles in this special section focus on data science and artificial intelligence for communication applications. The popularity of this Series continues to grow, attracting a lot of attention from both researchers and practitioners who are working to address various challenges in the network field through the advances of artificial intelligen...

The articles in this special section are dedicated to the application of artificial intelligence AI), machine learning (ML), and data analytics to address different problems of communication systems, presenting new trends, approaches, methods, frameworks, systems for efficiently managing and optimizing networks related operations. Even though AI/ML...

The articles in this special section present new technical advancements in applying AI and data science for communications.

The six papers in this special section address the application of artificial intelligence, machine learning, and data analytics at different layers and different applications of different types of communications networks. The objective of using these tools is the optimal design and improved operation of networks. These articles feature new opportun...

This is the second issue of the Data Science and Artificial Intelligence for Communications Series. The innovation in artificial intelligence (AI), machine learning (ML), and network data analytics provides a huge opportunity to revolutionize the world’s communications systems and user experience. Through gathering, processing, learning, and contro...

The five articles in this special section focus on data science and artificial intelligence (AI) for communications. The innovation in AI, machine learning (ML), and network data analytics provides a huge opportunity to revolutionize the world’s communications systems and user experience. Through gathering, processing, learning, and controlling the...

The articles in this special section focus on network testing and analytics. User-centric quality of experience (QoE) measurement has been attracting attention as a complement to existing network- centric measurement. Understanding the correlation between network quality of service (QoS) metrics and the user perceived QoE can drive improvement in s...

This Series continues to receive and report interesting works on network testing as well as network analytics. On network testing, we see increasingly sophisticated tools or even a framework of tools that can test beyond what existing commercial tools could do and render results with good insights. They have the potential to further push the state...

Network testing could have a wide spectrum. It is not confined to lab testing with all configurations preset and all results easily reproducible. It could go to field testing with a much larger and even operational network as the testbed. This could also be hybrid testing with parts of the testbed being real and the other parts being emulated or ev...

Welcome to the first issue of the updated IEEE Series on Network Testing and Analytics. The objective of this Series is to provide a forum across academia and the industry to address the application and development of a range of testing and analytics capabilities and practices that are required to model, measure, and evaluate the effectiveness of n...

In this paper we present a new resource management architecture, which provides end-to-end QoS guarantees to individual flows in a large, hierarchical network, such as the global Internet. This framework provides exact reservation and guaranteed resources to individual flows and, therefore, has the same expressive power as the IntServ model, while...

In this paper we present experimental results evaluating the performance and fairness of FAST TCP in a series of tests involving realistic low rate network access scenarios. Links both using the DOCSIS cable modem medium access control (MAC) cable modem and simple low rate links were investigated. We seek to compare our expectations from theory wit...

This paper addresses the problem of capacity pro-visioning in a QoS enabled IP network, subject to satisfying varying performance requirements for different traffic classes. These performance requirements are viewed in terms of mean end-to-end delays required for the various traffic classes or in terms of random variations of their delays (i.e., ji...

In the past, single service networks such as the PSTN were relatively simple to design and dimension in order to meet specific service objectives. The arrival of next generation networks (NGN) based upon IP technology and supporting multiple service types has meant greater challenges for network designers, managers and planners as customers may exp...

Traffic matrices constitute essential inputs in a wide variety of network planning and management functions as they provide the traffic volumes that flow between the node pairs in a network. In operational IP networks, it is desirable that traffic matrix (TM) estimation relies on information that is directly obtainable from SNMP system measurements...

In this paper, we examine the issues that surround IP network design with quality of service (QoS) guarantees and propose a new network design methodology. The proposed network design model takes account of the new QoS technologies (i.e., DiffServ/MPLS) and allows for multiple delay constraints so that guaranteed performance can be achieved for eac...

This technical report describes the process of implementing a testbed, which was used in [1] to experimentally evaluate the performance of FAST under typical 'edge of network' scenarios involving DOCSIS (Data Over Cable Service Interface Specification) cable system.

There is strong evidence that the efficiency of the Internet is limited by its existing TCP congestion control system. A replacement, FAST, has been shown to improve performance in high-speed networks. In order to achieve widespread acceptance and standardisation, it must also be tested in environments more typical of the existing Internet. This pa...

This paper considers the combined problem of optimal QoS partition and routing (problem QPQR-G) for a QoS framework in which a performance dependent cost function is associated with each network element and the QoS metric is additive (e.g. delay, jitter). This problem has been addressed in the context of unicast connections and multicast trees only...

This paper describes a network planning tool that has been developed in order to provide support for network planners who need to dimension next generation IP networks to meet quality of service (QoS) objectives. Specifically, the proposed network planning tool takes account of the new technologies that enable QoS in IP-based networks (i.e., DiffSe...

In this paper, we propose a dimensioning model that takes account of the new technologies that can provide QoS (i.e., DiffServ/MPLS) and allows for multiple delay constraints so that guaranteed performance can be achieved for each of the traffic classes. The proposed model incorporates procedures that allow correlations and burstiness of IP traffic...

Traffic demands in the Internet today include delay-sensitive traffic that require better than the standard "best effort" service provided by IP networks. As a result, various new technologies and mechanisms to support all types of traffic with various performance requirements are being developed and implemented. The traditional methods for capacit...

This technical report is an extended version of a pa-per [1], which addresses the problem of dimensioning links in a multiservice IP network subject to satisfying varying performance requirements for different traffic classes. These performance re-quirements are viewed in terms of mean end-to-end delays re-quired for the various traffic classes or...

Recently, a framework for the development of multiservice IP networks based on a mixture of DiffServ-based QoS management and MPLS-based traffic engineering has been defined. This approach is based on the aggregation of individual flows into classes and on the provisioning of QoS to the class instead of a single flow. With these developments, it is...

We consider IP over ATM networking scenario, widely deployed today, where Asynchronous Transfer Mode (ATM) is used as a backbone network to provide high-speed transport for the Internet Protocol (IP) traffic. Telecommunication network carriers and Internet Service Providers (ISPs), that have deployed ATM as their backbone networks, need to model an...

In this paper, we address the problem of dimensioning links in a multiservice IP network to satisfy the varying delay (QoS) constraints for different traffic classes. The delay (QoS) performance objectives are viewed in terms of mean delays required for the various traffic classes or in terms of random variations of their delays (e.g., jitter) or a...

The network design problem considered in this paper provides the basis for the solution of many inter-esting and practical engineering problems, of which one notable example is the link capacity expansion problem of existing communication networks. We are particularly interested in the possibility of using this problem as a subproblem when solving...

Traffic-based decomposition models en-compass procedures required for modelling of the ba-sic network operations of superposition, departure and splitting, arising due to the common sharing of the re-sources and routing decisions taking place in packet-switched networks. It is desirable to study such mod-els for Markovian Arrival Processes (MAPs),...