Analysis and measurement of performances in time-constrained communication networks

Project Coordinator

C. Narduzzi

Starting Date

January 2006

Ending Date

March 2008

Description

The use of communication networks, particularly of Ethernet, to realize measurement and control functions, is becoming widespread in several application fields of the industrial area. These functions require real-time behaviour, which implies very restrictive conditions on acceptable values for latency, as well as on operating times, that need to be specified in deterministic terms. The term real-time network usually denotes networks for which both software protocols and hardware devices are specifically developed, in order to provide similar or better performances than field buses, presently deployed in factories. Several companies offer real-time systems based on custom software and hardware standards; relevant examples are Profibus and EtherCAT. In a long-term perspective, real-time networks may well become an essential infrastucture in modern manufacturing companies.

Design and realization of a real-time network need to account for several factors that may eventually influence its implementation, either by determining the degree of complexity, or by placing constraints: for instance, topology can be restricted by the placement and accessibility of the plant where the network is going to be deployed; in more innovative systems, one of the design aims is the coexistence, within the same network structure, of real-time data flows with non real-time flows. Although the latter have less restrictive timing requirements, they employ part of the available resources and may still influence real-time performances.
Potential users of real-time network technologies do realise the need to qualify real-time network performances by defining appropriate quantitative parameters and suitable measurement methods. This has led to significant efforts in the study of the problem, at the standardization level, by IEC international technical committees.

Evaluation of network performances is obviously a main topic in the area of Information and Communication Technology (ICT), where it is generally considered in terms of quality of service (QoS) analysis. However, the features of data traffic in a real-time network differ remarkably from those of traffic in a common communication network. In fact, transfers usually involve moderate quantities of data, but are subject to rather restrictive timing specifications. As far as the characterisation of a real-time network is concerned, QoS parameters like latency and timing jitter are the most important, as they may directly affect the response times of control systems and the accuracy of time references for measured data. Indicators related to transport capacity, such as the throughput, are less relevant, since available bandwidth is usually abundant. ICT applications that more closely resemble real-time network features are those involving constant bit rate (CBR) data transfers, such as digital telephony and video streaming. Existing similarities allow to rely on research results available in the ICT literature as a useful starting point, but it should be emphasised that existing methods and criteria cannot just be simply adapted to suit the real-time network environment. It should be noticed in particular that measurements for ICT applications do not require a particularly high degree of accuracy, since they are mostly aimed at supporting admission control (measurement-based admission control) and optimal resource management. Conversely, measurement accuracy requirements in real-time networks are determined by the allowable timing tolerances, which may be as low as a few microseconds.

Ensuring that all activities within a real-time network share a commong timing reference requires that all nodes be synchronised. This problem is common to several application fields, among which distributed measurement systems, and has been broadly dealt with in the literature. However, timing accuracy is not always regarded as a critical issue: for instance, the performances of the standard Internet network timing prtocol (NTP) are adequate for environmental monitoring applications, since synchronisation to within some tens of milliseconds can be achieved. Telecommunication network monitoring requires measurements to be much better synchronised, but in this case requirements are typically satisfied by incorporating in the measuring instrument a GPS receiver as a source for the timing reference. This solution is much more expensive and hardly viable in an industrial environment. This evidences the need for further studies, which need to specifically consider efficient synchronisation methods for real-time networks. At the software level, IEEE 1588 is the reference protocol, although it may not necessarily be the best suited solution.

Characterisation of measuring instrumentation in this field is mainly concerned with the analysis of data acquisition procedures and of the algorithms which estimate the parameters of interest. The study of best practices in the field of communications networks can provide, also in this case, useful indications for research on real-time networks; it should be mentioned in particular, that a reference framework for Internet measurements has already been defined in the RFC 2330 document. It must be stressed that validation of instruments for network measurements necessarily requires the inclusion of simulation activities, which may be complemented by hardware/software emulations of parts of a real-time network. In fact, it is difficult to carry out experimental activities on-line when a network is operating, particularly when the response of the instrument to network anomalies or rare events needs to be tested. Two widely used simulation environments are OPNET and Network Simulator 2 (ns-2). The former is a commercial tool, whereas ns-2 has been developed as open-source software and is largely employed in network research. Cost considerations aside (although this is by no means a secondary issue), an advantage of ns-2 is that it favours the build-up of research communities through the shared development of the software modules required for simulation of the network environments of interest.
As far as the possible need for network emulation is concerned, at least one tool of potential interest for the activities of the proposed research project is freely available. NIST Net is an emulation tool developed by the US National Institute of Standards and Technology (NIST); it can replicate, in a controlled and repeatable way, the main features of a network in terms of response times, time jitter, possible packet losses, etc. An important feature is that the tool can be calibrated, allowing the accurate reproduction of the desired time patterns. It can be considered as a useful support for the generation of data traffic containing known impairments, which may allow the verification of the correct operation of diagnostic tools.

Pubblications

  • BENETAZZO L, GIORGI G., NARDUZZI C. On the Analysis of Communication and Computers Networks by Traffic Flow Measurement. In: Proc. IEEE Instrum. Meas. Technol. Sorrento, Italy, 24-27 April 2006, p. 212-217.

  • BENETAZZO L, GIORGI G., NARDUZZI C. Analisi statistiche di misure di traffico in reti di comunicazione. In: Atti del XXIII Congresso Nazionale Associazione GMEE 2006. L'Aquila, Italia, 11-13 Settembre 2006

  • GIORGI G., NARDUZZI C. Detection of Anomalous Behaviors in Networks from Traffic Measurement. In: Proc. IEEE Instrum. Meas. Technol. Sorrento, Italy, 24-27 April 2006, p. 330-335.

  • GIORGI G., NARDUZZI C. Analysis of Traffic Flow Measurement by Rate-Interval Curves. In: ACM International Conference Proceeding Series. Pisa, Italy, 11-13 October 2006, vol. 180.

  • BENETAZZO L, GIORGI G., NARDUZZI C. On the Analysis of Communication and Computer Networks by Traffic Flow Measurement. IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, vol. 56; p. 1157-1164.

  • BENETAZZO L, GIORGI G., NARDUZZI C, STELLINI M. Studio di un ambiente di prova per misure di networking. In: Atti del XXIV Congresso Nazionale Associazione GMEE 2007. Torino, 05-07 settembre 2007.

  • GIORGI G., NARDUZZI C. Ottimizzazione delle misure di intensit√† del traffico in rete. In: Atti del XXIV Congresso Nazionale Associazione GMEE 2007. Torino, 05-07 settembre 2007, TORINO: C.L.U.T. Editrice, p. 379-388.

  • GIORGI G., NARDUZZI C. A Study of Measurement-based traffic models for network diagnostics. In: 2007 IEEE Instrumentation and Measurement Technology Conference Proceedings. Varsavia, Polonia, May 01-03, 2007.

  • GIORGI G., NARDUZZI C (2007). Modeling and Simulation Analysis of PTP Clock Servo. In: IEEE Symposium on Precision Clock Synchronization for Measurement, Control and Communication ISPCS07. Vienna, Austria, 01-03 ottobre 2007, p. 155-161.

  • GIORGI G., NARDUZZI C. Detection of Anomalous Behaviors in Networks From Traffic Measurements. IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, vol. 57; p. 2782-2791.

  • GIORGI G., NARDUZZI C. A study of measurement-based traffic models for network diagnostics. IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, vol. 57; p. 1642-1650.

  • GIORGI G., NARDUZZI C. Rate-interval curves - A tool for the analysis and monitoring of network traffic. PERFORMANCE EVALUATION, vol. 65; p. 441-462.

  • GIORGI G., C. NARDUZZI. Uncertainty of Quantile Estimates in the Measurement of Self-Similar Processes. In: Workshop on Advanced Methods for Uncertainty Estimation Measurements, AMUEM 2008. Sardagna, Trento, Italy.

  • GIORGI G., C. NARDUZZI. Joint pdf Plots: Telling Beetles from Ants in Network Traffic Measurement by Cross-Domain Analysis. In: IEEE Instrum. Meas. Techn. Conference, I2MTC2008. Victoria VancouverIsland, Canada, 12-15 May 2008.

  • GIORGI G., C. NARDUZZI, M. STELLINI. A Test Bed for Synchronization in Heterogeneous Network Environments. In: Proc. IEEE Instrum. Meas. Techn. Conference, I2MTC2008. Victoria, VancouverIsland, Canada, 12-15 May 2008.

Tools