Wojciech Molisz

Survivability function - a measure of disaster-based routing performance

The explosive growth of data traffic imposes critical requirements on core network survivability. Developments in wavelength-division multiplexing have strengthened this need. Survivability becomes increasingly crucial, since large traffic volumes are multiplexed onto a single fiber. A single cable cut can affect incredibly large groups of users, leading to catastrophic socioeconomic effects. This paper defines the network survivability function - the probability function of the percentage of total data flow delivered after failure and survivability attributes - the expected percentage of total data flow delivered after failure, the respective p-percentile values, the worst case survivability. Models for finding these survivability measures are described. The main goal in this paper is to investigate the survivability function for typical routing protocols used in the IP networks. Examples of survivability assessment of a typical wide area network employed in Poland illustrate the proposed approach.

Region protection/restoration scheme in survivable networks

In this paper we propose the novel concept of a region protection/ restoration, where one backup path protects a certain region of an active path. We show that using the region protection/restoration we can keep both restoration times and network resource utilization ratio at the reasonable level. Since the optimization problem of finding working and backup paths is NP-complete, we developed the heuristic algorithm. We show that in the worst case our algorithm gave network resource utilization ratio only about 3.9 percent higher compared to the optimal one returned by the CPLEX program. Results of the U.S. Long-Distance Network modeling show that region protection gives a good compromise between path and link protection.

A new approach to provide the differentiated levels of network survivability under a double node failure

In this paper we propose a class-based protection algorithm providing fast service recovery in WDM networks in case of a simultaneous failure of two network elements. We focus on providing various levels of service survivability in order to correspond to differentiated requirements of end-users. We show that assuring the service survivability in case of a double failure requires up to 40% more total link capacity, compared to the results for protection against a single node failure. The time of service recovery is also increased. However, the values of service recovery time may be reduced by limiting the scope of backup path protection.

A novel class-based protection algorithm providing fast service recovery in IP/WDM networks

In this paper we consider a multilayer architecture: IP-MPLS over optical transport networks (OTNs). Nodes have integrated functionality of optical cross connects (OXCs) in the optical layer and of IP routers in the IP layer. Any two IP routers can be connected together by an all-optical wavelength-division multiplexing (WDM) channel, called a lightpath. Wavelength converters may be installed in each node. We propose a class-based method providing fast service restoration and reducing the number of service recovery actions in the IP layer. We focus on service restoration time values and we show, that it is possible to decrease them, by reducing the size of active path protected area. This may be achieved for the price of link capacity utilization degradation.

Reliable routing and resource allocation scheme for hybrid RF/FSO networks

Significant success of wireless networks in the last decade has changed the paradigms of communication networks design. In particular, the growing interest in wireless mesh networks (WMNs) is observed. WMNs offer an attractive alternative to conventional cable infrastructures, especially in urban areas, where the cost of new installations is almost prohibitive. Unfortunately, the performance of WMNs is often limited by the cluttered RF spectrum and channel interference. Free space optical (FSO) technology is a possible solution for IP traffic connectivity, but weather-based disruptions, e.g., clouds, fog, and snow, highly affect the service availability in the network. However, FSO links can provide much higher data rates than RF links and do not interfere with RF links. In this paper, we first outline the important aspects that should be taken into account when determining allocation of network resources (link capacities) for the purpose of serving the demands. The latter part of the paper is to introduce the respective ILP model as well as heuristic algorithm to determine routing in hybrid RF/FSO networks with special focus on resistance of FSO links to weather-based disruptions.

Impact of WDM network topology characteristics on the extent of failure losses

In this paper we evaluate the influence of the network topological characteristics on the extent of losses after attacks. For that purpose, we use the ATRF (attack-to-random failure) multiplier. Modeling results show that when using a standard metrics of distance in path computations, the number of connections broken due to attacks on irregular networks is significantly greater, compared to random failures. This is true independent of the algorithm used for path finding.

On certain behavior of scale-free networks under malicious attacks

This paper evaluates performance of scale-free networks in case of intentional removal of their nodes. The distinguishing feature of this kind of networks (Internet is an excellent example) is the power law distribution of node degrees. An interesting behavior of scale-free networks, if node removal process is performed sufficiently long, is manifested by their migration to random networks. The main idea of our research is to quantify this process. In contrast to well explored parameters like: characteristic path length or clustering coefficient, we propose the new ones: mean maximum flow, centre of gravity of node degree distribution and other. To the best of our knowledge, these measures are proposed for the first time. Our results confirm that the migration process steps relatively fast.

Global communications newsletter: June 2013

Being the largest academic and industrial center as well as the famous tourist destination at the Croatian Adriatic coast, the city of Split was a perfect place to start a conference on software, telecommunications and computer networks. The SoftCOM conference started in 1993 as a joint project of the University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Croatian Telecom and World Trade Center Split.

Quality of resilience in IP-based Future Internet communications

Future Internet (Next Generation Network, or Internet 3.0) is a further step in evolution of the Internet. The research activities related to Future Internet cover a wide spectrum of topics including: network architectures (protocols), network technologies, applications, etc. NGN assumes one basic network protocol (IPv6), Quality of Service (QoS) provided by IP-based infrastructure, improved signaling, improved security and other functionalities. In this paper we assume the utilization of Differentiated Services (DiffServ) functionality. In particular, we propose a model of providing the differentiated levels of service resilience. Such solution seems reasonable, since not every service requires the same level of protection. In particular, we concentrate on protecting the flows against failures being the result of malicious human activities. Results of simulation provide a compromise between the amount of capacity needed to guarantee protection, and the number of connections broken after failures.

ƒ-cycles — a new approach to providing fast service recovery at low backup capacity overhead

In this paper we propose the class-based protection algorithm providing fast service recovery in optical networks using the new concept of fast protection cycles (f-cycles). We focus on service restoration time values and we show, that it is possible to decrease them, by reducing the size of active path protected area. We prove that it is also possible to reduce the ratio of link capacity utilization by applying the concept of p-cycles.