Mikel Izal

On the influence of self-similarity on optical burst switching traffic

In this paper we provide a characterization of OBS traffic (burst size, interarrival time and scaling behavior) when the input traffic is long-range dependent The analysis shows that the influence of self-similarity on blocking probability is negligible, since the arrival process can be assumed to be Poisson in the timescale of interest for burst blocking. However, the impact for optical buffer dimensioning is significant. On the other hand, the scaling region is shifted to larger timescales while traffic variability at low timescales is increased. These findings serve to accurately dimension number of output ports and optical buffers in OBS routers when the incoming traffic comes from a large population of Internet users.

The European Traffic Observatory Measurement Infrastructure (ETOMIC): a testbed for universal active and passive measurements

The European Traffic Observatory is a European Union VI Framework Program sponsored effort, within the Integrated Project EVERGROW, that aims at providing a paneuropean traffic measurement infrastructure with high-precision, GPS-synchronized monitoring nodes. This paper describes the system and node architectures, together with the management system. On the other hand, we also present the testing platform that is currently being used for testing ETOMIC nodes before actual deployment.

The European traffic observatory measurement infrastructure (ETOMIC)

The European traffic observatory is a European Union VI framework program sponsored effort, within the integrated project EVERGROW, that aims at providing an panEuropean traffic measurement infrastructure with high-precision, GPS-synchronized monitoring nodes. This paper describes the system and node architectures, together with the management system.

Blocking time analysis of OBS routers with arbitrary burst size distribution

The blocking time distribution for an OBS router is obtained, under the assumption of Poisson-arriving bursts with Pareto, Gaussian and exponential burst size distributions. Analytical expressions are provided as a function of number of wavelengths per port. Such expressions can be used to dimension fiber delay lines (FDLs) and to perform end-to-end delay estimation. On the other hand, we show that the blocking time distribution becomes exponential as the number of wavelengths increases, regardless of the burst size distribution. Since the burst size distribution is determined by the burst assembly algorithm at the network edges, we conclude that the burst assembly algorithm have no influence on both burst blocking probability and burst blocking time in future DWDM networks.

Analysis of Internet services in IP over ATM networks

This paper presents a trace-driven analysis of IP over ATM services from a user-perceived quality of service standpoint. QoS parameters such as the sustained throughput for transactional services and other ATM layer parameters such as the burstiness (MBS) per connection are derived. On the other hand, a macroscopic analysis that comprises percentage of flows and bytes per service, TCP transaction duration and mean bytes transferred in both directions is also presented. The traffic trace is obtained with novel measurement equipment that combines header extraction hardware and a high-end UNIX workstation capable of providing a timestamp accuracy in the order of microseconds. The ATM link under analysis concentrates traffic from a large population of 1500 hosts from the Public University of Navarra campus network, that produced 1700000 TCP connections approximately in the measurement period of one week. The results obtained from such a wealth of data suggest that QoS is primarily determined by transport protocols and not by ATM bandwidth. The sustained throughput of TCP connections never grows beyond 80 Kbit/s with 70% probability in the data transfer phase (i.e., in the ESTABLISHED state) and we observe a strong influence of the connection establishment phase in the user-perceived throughput. On the other hand, the burstiness of individual TCP connections is rather small, namely TCP connections do not produce bursts according to the geometric law given by slow start and commonly assumed in previously published studies.

Delay-throughput curves for timer-based OBS burstifiers with light load

The optical burst switching (OBS) burstifier delay-throughput curves are analyzed in this paper. The burstifier incorporates a timer-based scheme with minimum burst size, i.e., bursts are subject to padding in light-load scenarios. Precisely, due to this padding effect, the burstifier normalized throughput may not be equal to unity. Conversely, in a high-load scenario, padding will seldom occur. For the interesting light-load scenario, the throughput-delay curves are derived and the obtained results are assessed against those obtained by trace-driven simulation. The influence of long-range dependence and instantaneous variability is analyzed to conclude that there is a threshold timeout value that makes the throughput curves flatten out to unity. This result motivates the introduction of adaptive burstification algorithms, which provide a timeout value that minimizes delay, yet keeps the throughput very close to unity. The dependence of such optimum timeout value with traffic long-range dependence and instantaneous burstiness is discussed. Finally, three different adaptive timeout algorithms are proposed, which trade off complexity versus accuracy.

OBS network model for OMNeT++: a performance evaluation

Optical Burst Switching (OBS) is an optical switching technology capable of supporting large demands for bandwidth in optical backbones with Wavelength Division Multiplexing (WDM). This paper presents an OBS simulation model for the discrete event simulator OMNeT++. The performance of this model is compared with the performance of the well-known INET simulation model for IP networks. Both models show similar performance results. The OBS model is faster but uses more dynamic memory.

Techniques for better alias resolution in Internet topology discovery

One of the challenging problems related with network topology discovery in Internet is the process of IP address alias identification. Topology information is usually obtained from a set of traceroutes that provide IP addresses of routers in the path from a source to a destination. If these traceroutes are repeated between several source/destination pairs we can get a sampling of all IP addresses for crossed routers. In order to generate the topology graph in which each router is a node, it is needed to identify all IP addresses that belong to the same router. In this work we propose improvements over existing methods to obtain alias identification related mainly with the types and options in probing packets.

ETOMIC Advanced Network Monitoring System for Future Internet Experimentation

ETOMIC is a network traffic measurement platform with high precision GPS-synchronized monitoring nodes. The infrastructure is publicly available to the network research community, supporting advanced experimental techniques by providing high precision hardware equipments and a Central Management System. Researchers can deploy their own active measurement codes to perform experiments on the public Internet. Recently, the functionalities of the original system were significantly extended and new generation measurement nodes were deployed. The system now also includes well structured data repositories to archive and share raw and evaluated data. These features make ETOMIC as one of the experimental facilities that support the design, development and validation of novel experimental techniques for the future Internet. In this paper we focus on the improved capabilities of the management system, the recent extensions of the node architecture and the accompanying database solutions.

Evaluation of preemption probabilities in OBS networks with burst segmentation

Preemption techniques have been recently proposed for service differentiation in optical burst switching (OBS) networks. According to V. Vokkarane et al. (2003), an incoming burst with the same priority that the burst in service will preempt the wavelength if the residual length of the burst in service is smaller than the incoming burst transmission time. For a network scenario with no wavelength conversion, the preemption probability is evaluated assuming exponential, Gaussian and Pareto-distributed burst sizes. Knowledge of the preemption dynamics at an OBS switch is a fundamental issue in performance evaluation, since the downstream switches will surely be affected. An analytical upper bound is provided, that shows that the preemption probability depends on the burst size distribution, which in turn depends on the burst assembly technique used at the network edges. On the other hand, not only truncated bursts result from preemption, as reported in other studies, but also the burst size distribution for preempting bursts is shifted to larger values.