Markus Laner

Users in cells: A data traffic analysis

We present a large-scale cell based measurement analysis of the user behavior in a live operational HSDPA network. The motivations are: first, to understand the statistical properties of users in cells for refining network planning procedures; and second, to provide realistic traffic models for simulations of cellular packet-oriented networks. We provide an analysis of mean cell load over daytime, as well as models for short-term cell load in terms of user activity and throughput. Furthermore, an evaluation of user-sessions with respect to duration and mean throughput is given. Our findings lead to four different models reflecting realistic user-traffic load of cellular networks. To verify the concept we present respective simulations which investigate multiuser-scheduling in an LTE-network. They show that conventional simulation settings can lead to an overestimation of performance.

Dissecting 3G uplink delay by measuring in an operational HSPA network

Users expect mobile Internet access via 3G technologies to be comparable to wired access in terms of throughput and latency. HSPA achieves this for throughput, whereas delay is significantly higher. In this paper we measure the overall latency introduced by HSUPA and accurately dissect it into contributions of USB-modem (UE), base station (NodeB) and network controller (RNC). We achieve this by combining traces recorded at each interface along the data-path of a public operational UMTS network. The actively generated sample traffic covers real-time applications. Results show the delay to be strongly dependent on the packet size, with random components depending on synchronization issues. We provide models for latency of single network entities as well as accumulated delay. These findings allow to identify optimum settings in terms of low latency, both for application and network parameters.

Time synchronization performance of desktop computers

Time synchronization is a vital requirement for various applications. Especially the synchronization of desktop computers to the Coordinated Universal Time (UTC) yields numerous use cases, such as distributed measurements. Several solutions address this need, at different levels of price and accuracy. In this work we evaluate on the achievable precision in time synchronization of a desktop PC, for example, assisted by a low-budget GPS receiver. This is achieved by a novel measurement setup, which is comparing the software synchronized internal clock of the PC to a rubidium frequency standard. Our results show, that the synchronization offsets of the software clocks of all tested PCs have positive mean (time lag) in the order of 10 µs. The respective standard deviation is typically an order of magnitude lower. Thereby the unknown interrupt latency is the limiting factor for the accuracy. With this work we show that today (2011) 10 µs of precision can be achieve at very low cost.

Modeling randomness in network traffic

A continuous challenge in the field of network traffic modeling is to map recorded traffic onto parameters of random processes, in order to enable simulations of the respective traffic. A key element thereof is a convenient model which is simple, yet, captures the most relevant statistics. This work aims to find such a model which, more precisely, enables the generation of multiple random processes with arbitrary but jointly characterized distributions, auto-correlation functions and cross-correlations. Hence, we present the definition of a novel class of models, the derivation of a respective closed-form analytical representation and its application on real network traffic. Our modeling approach comprises: (i) generating statistical dependent Gaussian random processes, (ii) introducing auto-correlation to each process with a linear filter and, (iii) transforming them sample-wise by real-valued polynomial functions in order to shape their distributions. This particular structure allows to split the parameter fitting problem into three independent parts, each of which solvable by standard methods. Therefore, it is simple and straightforward to fit the model to measurement data.

Packet delay measurements in reactive IP networks

Discussions of how to measure the performance of computer networks for various applications have been ongoing for over twenty years in the area of network research. The continual increase of data traffic volume has reached a point at which solving any network problem by over-provisioning is not suitable. The quest for alternatives makes it vital to have well-defined metrics for evaluating and sustaining the performance of networks.

Outer-Loop Power Control in a live UMTS network: Measurement, analysis and improvements

The Outer Loop Power Control (OLPC) in WCDMA systems allows to achieve a defined Quality of Service (QoS) for every link. This reduces resources and interference, hence, increases the system-wide throughput. In this paper we present uplink OLPC related large-scale measurements, performed on live Iub-interfaces. Evaluations of the actual implemented algorithm show that it converges slowly; the reason being that the QoS is estimated by CRC. As the uncoded Bit Error Ratio (BER) holds information about the QoS, this parameter can be used to increase convergence speed of the OLPC. We present a statistical model of the control path of the OLPC which takes the uncoded BER information into account. Additionally, we propose a new OLPC algorithm that shows fast convergence in simulations, resulting in a reduction of 0.2 dB to 1 dB of the mean required signal-to-interference ratio, depending on the user mobility.

Parsimonious Network Traffic Modeling By Transformed ARMA Models

Generating synthetic data traffic, which statistically resembles its recorded counterpart is one of the main goals of network traffic modeling. Equivalently, one or several random processes shall be created, exhibiting multiple prescribed statistical measures. In this paper, we present a framework enabling the joint representation of distributions, autocorrelations and cross-correlations of multiple processes. This is achieved by so called transformed Gaussian autoregressive moving-average models. They constitute an analytically tractable framework, which allows for the separation of the fitting problems into subproblems for individual measures. Accordingly, known fitting techniques and algorithms can be deployed for the respective solution. The proposed framework exhibits promising properties: 1) relevant statistical properties such as heavy tails and long-range dependences are manageable; 2) the resulting models are parsimonious; 3) the fitting procedure is fully automatic; and 4) the complexity of generating synthetic traffic is very low. We evaluate the framework with traced traffic, i.e., aggregated traffic, online gaming, and video streaming. The queueing responses of synthetic and recorded traffic exhibit identical statistics. This paper provides guidance for high-quality modeling of network traffic. It proposes a unifying framework, validates several fitting algorithms, and suggests combinations of algorithms suited best for specific traffic types.

A benchmark methodology for end-to-end delay of reactive mobile networks

Modern mobile communication networks handle user traffic in terms of data streams rather than single packets. Resources are allocated per stream as a reaction on its history. In particular, the networks delay figure depends on the entire traffic pattern injected by the user. This work aims to find a comprehensive measure for the delay behavior of reactive networks. A novel measurement methodology is introduced which enables the fast assessment of the delay performance for a multitude of traffic patterns. The results are condensed to a handful of typical classes by simultaneously minimizing the loss of information. Those classes enable an intuitive representation of the networks latency behavior. Further, they indicate which patterns are preferable for fair benchmarking. We demonstrate the practicability of the approach by benchmarking operational networks, namely, LTE (non-reactive) and HSPA (reactive). The analysis shows that the proposed method works well for both technologies, outperforming standard methods which a priori assume non-reactiveness of the network under test.

Measurement Aided Model Design for WCDMA Link Error Statistics

Accurate models for error characteristics of radio channels are essential to estimate performance of data transmission. This paper analyzes error-gap and error-burst statistics of the WCDMA dedicated channel (DCH) and provides respective models. We start with an analytical study of the outer-loop power control mechanism (OLPC), which gives directions to this work, revealing that the OLPC strongly influence error behavior. We prove this by large-scale measurements at the Iub-interface of a live UMTS network. Beside statistical evaluation of the measurement data we provide a simple generative hidden Markov model for emulation of DCH errors. It is able to characterize any arbitrary DCH by only two parameters, with an accuracy below 1%, in terms of Kullback-Leibler divergence. This novel model presents an accurate light-weight alternative to complex tools for simulation of WCDMA DCH connections. It also covers all other communication technologies deploying a similar power control mechanism.

Traffic models for machine-to-machine (M2M) communications: types and applications

Abstract Machine-to-machine (M2M) or machine-type communication (MTC) is expected to have a significant traffic share in future wireless networks. It exhibits considerably different traffic patterns than human-type communication and, thus, requires new traffic models and simulation scenarios. Such models should (i) accurately capture the behavior of a single MTC device and (ii) enable the concurrent simulation of massive numbers of devices (e.g., up to 30,000 devices per cell) with their potential synchronous reactions to an event. In general, source traffic models (i.e., each device is modeled as an autonomous entity) provide higher precision and flexibility. However, their complexity grows quadratically with the number of devices. Aggregated traffic models, on the contrary, are far less precise but their complexity is mainly independent of the number of devices. This chapter presents several modeling strategies, namely: (i) aggregated traffic, (ii) source traffic, and (iii) a hybrid approach. The three models are explained and compared through a common use case. It allows both to illustrate the trade-off between accuracy and complexity and to guarantee the comparability of future studies by the deployment of common models.