Poul E. Heegaard

Network survivability modeling

Critical services in a telecommunication network should be continuously provided even when undesirable events like sabotage, natural disasters, or network failures happen. It is essential to provide virtual connections between peering nodes with certain performance guarantees such as minimum throughput, maximum delay or loss. The design, construction and management of virtual connections, network infrastructures and service platforms aim at meeting such requirements. In this paper we consider the networks ability to survive major and minor failures in network infrastructure and service platforms that are caused by undesired events that might be external or internal. Survive means that the services provided comply with the requirement also in presence of failures. The network survivability is quantified as defined by the ANSI T1A1.2 committee which is the transient performance from the instant an undesirable event occurs until steady state with an acceptable performance level is attained. The assessment of the survivability of a network with virtual connections exposed to link or node failures is addressed in this paper. We have developed both simulation and analytic models to cross-validate our assumptions. In order to avoid state space explosion while addressing large networks we decompose our models first in space by studying the nodes independently and then in time by decoupling our analytic performance and recovery models which gives us a closed form solution. The modeling approaches are applied to both small and real-sized network examples. Three different scenarios have been defined, including single link failure, hurricane disaster, and instabilities in a large block of the system (transient common failure). The results show very good correspondence between the transient loss and delay performance in our simulations and in the analytic approximations.

Differentiated Availability in Cloud Computing SLAs

Cloud computing is the new trend in service delivery, and promises large cost savings and agility for the customers. However, some challenges still remain to be solved before widespread use can be seen. This is especially relevant for enterprises, which currently lack the necessary assurance for moving their critical data and applications to the cloud. The cloud SLAs are simply not good enough. This paper focuses on the availability attribute of a cloud SLA, and develops a complete model for cloud data centers, including the network. Different techniques for increasing the availability in a virtualized system are investigated, quantifying the resulting availability. The results show that depending on the failure rates, different deployment scenarios and fault-tolerance techniques can be used for achieving availability differentiation. However, large differences can be seen from using different priority levels for restarting of virtual machines.

Ant system for service deployment in private and public clouds

Large-scale computing platforms that serve thousands or even millions of users through the Internet are on a path to become a pervasive technology available to companies of all sizes. However, existing technologies to enable this kind of scaling are based on a hierarchically managed approach that does not scale equally well. Moreover, existing systems are also not equipped to handle the dynamism that may emerge as a result of severe failures or load surges. In this paper, we conjecture that using self-organizing techniques for system (re)configuration can improve both the scalability properties of such systems as well as their ability to tolerate churn. Specifically, the paper focuses on deployment of virtual machine images onto physical machines that reside in different parts of the network. The objective is to construct balanced and dependable deployment configurations that are resilient. To accomplish this, a method based on a variant of Ant Colony Optimization is used to find efficient deployment mappings for a large number of virtual machine image replicas that are deployed concurrently. The method is completely decentralized; ants communicate indirectly through pheromone tables located in the nodes. An example scenario is presented and simulation results are obtained for the method.

Cost-efficient deployment of collaborating components

We study the problem of efficient deployment of software components in a service engineering context. Run-time manipulation, adaptation and composition of entities forming a distributed service is a multi-faceted problem challenged by a number of requirements. The methodology applied and presented can be viewed as an intersection between systems development and novel network management solutions. Application of heuristics, in particular artificial intelligence in the service development cycle allows for optimization and should eventually grant the same benefits as those existing in distributed management architectures such as increased dependability, better resource utilization, etc. The aim is finding the optimal deployment mapping of components to physically available resources, while satisfying all the non-functional requirements of the system design. Accordingly, a new component deployment approach is introduced utilizing distributed stochastic optimization.

Efficient simulation of network performance by importance sampling

Simulation is a flexible means for assessment of the quality of service offered by a telecommunication system. However, when very strict requirements are put on the quality of service, the simulatio ...

Empirical observations of traffic patterns in mobile and IP telephony

This paper provides recent empirical traffic data and observations of telephony traffic patterns in mobile and IP telephony. These are compared with old telephony patterns from Public Switched Telephone Networks (PSTN) to investigate potential evolution and impact on traffic characterisations due to technology changes from fixed to mobile phones, changes in quality from fixed phone to mobile and IP telephone, changes in tariffs from usage based to flat-rate subscriptions, and appearance of alternative message based communication means. The results show different daily and weekly traffic profiles compared to PSTN telephony. In particular, the profile of international calls is significantly changed. Furthermore, the average call holding times show significant variations over the day in flat-rate subscriptions. Finally, the results indicate that the Short Message Service (SMS) seems to serve as a supplement to phone calls, in particular in the evenings, which might change call holding time distribution and traffic intensities.

Survivability quantification of communication services

Our society is heavily dependent on a wide variety of communication services. These services must be available even when undesirable events like sabotage, natural disasters, or network failures happen. The network survivability as defined by the ANSI T1A1.2 committee is the transient performance from the instant an undesirable event occurs until steady state with an acceptable performance level is attained. In this paper we assess the survivability of a network with virtual connections exposed to link or node failures. We have developed both simulation and analytic models to cross validate our assumptions. In order to avoid state space explosion while addressing large networks we decompose our models first in space by studying the nodes independently and then in time by decoupling our analytic performance and recovery models which gives us a closed form solution. The modeling approaches are applied to two network examples. The results show very good correspondence between the transient loss and delay performance in our simulations and in the analytic approximations.

Self-tuned refresh rate in a swarm intelligence path management system

CE-ants (Cross Entropy ants) is a distributed, robust and adaptive swarm intelligence system for dealing with path management in communication networks. This paper focuses on strategies for adjusting the overhead generated by the CE-ants as the state of the network changes. The overhead is in terms of number of management packets (ants) generated, and the adjustments are done by controlling the generation rate of ants traversing the network. The self-tuned strategies proposed in this paper detect state changes implicitly by monitoring parameters and ant rates in the management system. Rate adaptation is done both in the network nodes and in the peering points of the virtual paths. The results are promising, and compared to fixed rate strategies the self-tuned strategies show a significant saving (70-85%) in number of packets, and has similar (even slightly better) data packet delay and service availability. The rate adaptation in network nodes provides fast restoration with short path detection times and hence also high service availability. The implicit self-tuned ant rate in the path endpoints improves the convergence time on link state events without flooding the network with management packets in steady state when these are not required.

Overhead reduction in a distributed path management system

CEAS (cross entropy ant system) is a distributed, robust and adaptive swarm intelligence system for path management in communication networks. This paper focuses on strategies for handling the overhead in terms of processing cycles, memory storage, and number of management packets (ants) generated by CEAS when the state of the network changes. Pheromone sharing is introduced such that virtual connections with common sub-paths are sharing information and cooperate in the path finding when the paths have the same destination and the same objective function. The sharing of information reduces the required memory in each node significantly on the expense of an increase in the size of the management packets. However, the packets are still rather small. The cooperation also leads to an improvement in convergence rates which again results in reduced transmission overhead. A rate adjustment scheme is also proposed. The scheme is self-tuned and detects state changes implicitly and sets packet rates accordingly by monitoring parameter values in the management system. Rate adaptation can be done both in the network nodes and at the end-points of a virtual path. Compared to a fixed rate strategy the self-tuned strategies show a significant reduction in the number of packets generated, while maintaining the same data packet delay and service availability level. The self-tuned rate adjustment in the network nodes provides fast restoration with short path detection times, which ensures high service availability. The self-tuned ant rate in the end-points avoids flooding the network with management packets when these are not required. The performance and overhead of CEAS are compared to those of the link state routing currently in use in todays networks. The results show that CEAS outperforms link state routing both with respect to performance and overhead when the network experiences transient link failures, while the opposite is the case with long lived failures.

Group Based Traffic Shaping for Adaptive HTTP Video Streaming by Segment Duration Control

Adaptive video streaming is based on the concept of allowing the quality level of a video stream to change during its lifetime based on certain parameters. This approach makes the video stream more tolerant with respect to fluctuations in available bandwidth during the session lifetime. In order for a video stream to decide which quality level to request from the server, it is typical to use both parameters from the client side such as CPU load and estimated parameters from the network side such as available bandwidth. In order for available bandwidth estimations to become more accurate, it is beneficial with some degree of traffic shaping. This paper describes a new method of achieving a traffic shaping effect for adaptive video streaming sessions delivered to members of a specific user group identified by their unique network destination. The novelty of the method is represented by the use of segment size control per video session, rather than active shaping of each session. The effect of the method is analyzed by means of measuring packet IAT and video segment fetch durations, in a controlled lab environment using the Smooth Streaming framework from Microsoft.