2742 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 12, DECEMBER 2008 One-Way Delay Measurement: State of the Art Luca De Vito, Sergio Rapuano, Member, IEEE, and Laura Tomaciello Abstract—Nowadays, the evaluation of performance measure- ment in computer networks is an important issue. To ensure the quality of service of the network communication, one of the most important network performance parameters is the one-way delay (OWD). For accurate OWD estimation, it is essential to consider some parameters that can influence the measure, such as the operating system and, in particular, the threads, which are concurrent with the measurement application. Moreover, OWD estimation is not an easy task, because it can be affected by synchronization uncertainties. This paper aims to review the different solutions proposed in the scientific literature for OWD measurement. These solutions adopt different methods to guarantee a reasonable clock synchronization based on the Network Time Protocol, the Global Positioning System, and the IEEE 1588 Standard. These different approaches are critically reviewed, showing their advantages and disadvantages. Index Terms—Global Positioning System (GPS), IEEE 1588, network, Network Time Protocol (NTP), one-way delay (OWD), synchronization. I. I NTRODUCTION A S COMPUTER networks become more complex and larger, measurement infrastructures and methodologies become essential in characterizing network performances [1]. The metrics of the greatest relevance for network performance can be divided into four main groups: 1) availability; 2) loss and error; 3) delay; and 4) bandwidth. Availability metrics assess how robust the network is, i.e., the percentage of time the network is running without any problems that impact the availability of services. Loss and error metrics indicate network congestion conditions, transmission errors, and/or equipment malfunctioning. They usually measure the fraction of packets lost in a network due to buffer overflows or the fraction of errored bits or packets. Bandwidth metrics assess the amount of data that a user can transfer through the network in a time unit that is both dependent and independent of the existing network traffic. Finally, delay metrics also assess network congestion conditions or the effect of routing changes. They measure the delay [one-way delay (OWD) and round- trip delay (RTD)] and Internet Protocol delay variation (IPDV, Manuscript received October 12, 2007; revised April 23, 2008. First pub- lished June 13, 2008; current version published November 12, 2008. L. De Vito and L. Tomaciello are with the Laboratory of Signal Process- ing and Measurement Information, Department of Engineering, University of Sannio, 82100 Benevento, Italy, and also with the Benevento Research Labo- ratory, Telsey Telecommunications S.p.A, 82018 San Giorgio del Sannio, Italy (e-mail: devito@unisannio.it; laura.tomaciello@unisannio.it). S. Rapuano is with the Department of Engineering, University of Sannio, 82100 Benevento, Italy (e-mail: rapuano@unisannio.it). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIM.2008.926052 or “jitter”) of the packets transferred by a network [2]. In particular, the OWD is the time between the occurrence of the first bit of a packet on the first observation point, e.g., the transmitting monitor interface, and the occurrence of the last bit of a packet on the second observation point (RFC 2679 [3]). The RTD is considered to be the time interval between the time instant a request packet is sent by a source node and the time instant a response packet is received from the destination node (see RFC 2681 [4]). Finally, the IPDV is the difference in the OWD of a selected pair of packets in a test stream (see RFC 3393 [5]). It is worth noting that each metric deals with time: time percentage, time delay, and time unit; for this reason, the main indicators in evaluating the network performance are the network delays. Network delays are composed of three components: 1) equipment delay; 2) transmission delay; and 3) propagation delay [6]. The first is the delay introduced by the equipment before it becomes emitting equipment. This delay consists of the processing time, packet switching, and queueing delays and depends on the network load and congestion. The second is the time taken to transmit all the bits of the frame containing a packet. It depends on the data rate, media, and distance and can only be controlled in a limited way by the network planners [7]. The third is the time between the emission of the first bit (or the last bit) of a packet by the transmitting equipment and the reception of this bit by the receiving equipment [8]. Therefore, the equipment, transmission, and propagation delay indicators allow the determination of the time that the packet spends to travel from source to destination. This time is called OWD. The OWD, as shown in Fig. 1, is constituted by three time contributions. Fig. 1(a) represents the equipment delay, which is the time interval between instant t 0 , when the packet is scheduled for sending, and instant t 1 , when the packet reaches the interface. Fig. 1(b), which represents the transmission delay, is the time interval between instants t 1 and t 2 , when the packet is completely transmitted onto the medium. Finally, Fig. 1(c) in- dicates the propagation delay, which is the time interval from t 2 up to t 3 , when the packet reaches the destination interface [9]. The OWD can be obtained by the sum of the transmission delay and the propagation delay, which could be defined as the time between the emission of the first bit of a packet by the source and the reception of the last bit of this packet by the receiver [3]. Some papers [10]–[24] deal with a reliable and accurate way to measure the OWD. To measure the OWD, a sequence of probe packets is to be sent from one end of the monitored network to the other end. Each probe packet is marked with 0018-9456/$25.00 © 2008 IEEE