Traffic characterization and LTE performance analysis for M2M communications in smart cities

Abstract

The paper presents a model for the characterization of Machine-to-machine (M2M) traffic and the performance evaluation of LTE access to support M2M communication, embedded into a web-based application. The application enables the study of the traffic produced by realistic M2M elements in the context of smart cities. Packet generation for each machine is modeled by means of three mathematical distributions (Poisson, Beta, and Deterministic), which makes it possible to represent a wide variety of M2M applications. A case study was constructed based on the city of Montreal. Real data on the position of machines were retrieved from public datasets. The case study includes a realistic representation of the LTE infrastructure that allows the estimation of traffic load at each eNodeB, as well as other performance indexes, such as collision probability and access delay. The use of real geographic information enables visual analyses aiming at identifying bottlenecks and possible roadblocks to the M2M integration in the LTE infrastructure.

Introduction

M2M communications are usually referred to as the transmission of information among smart objects (i.e., devices with communication capabilities), without any human intervention. The importance of this kind of communication has risen in the last decade, due to the steep increase in the number of devices that are able to autonomously communicate with each other. The growth will accelerate in the next future, and the number of communicating devices is expected to rise to several billions in 2020. The connection of this large amount of communicating objects, also known as Internet of Things (IoT), needs to be supported by one of the existing communication infrastructures. LTE is currently seen as one of the best candidates due to its ubiquitous access and large spectrum availability. Moreover, 3GPP has – in Release 13 in 2014 – proposed Narrow Band IoT (NB-IoT), a new LTE-protocol specifically conceived to provide low power access to a large number of devices. The small transmission power and the reduced bandwidth (i.e., 200 kHz), makes NB-IoT particularly suitable for M2M communications.

The LTE infrastructure was, however, originally conceived to support human traffic, that is sensitively different from M2M traffic. First of all, M2M traffic is mostly in the uplink direction, whereas Human Type Communications (HTC) are most frequently distributed in the downlink. Second, the packet size in M2M applications is essentially in the order of hundreds of bits, considerably smaller than in HTC. Third, M2M devices are predominantly¹ installed in fixed locations and do not require mobility, which is often needed in HTC. Moreover, human traffic is strongly correlated to human activity time (e.g., busy during the day and sleeping at night), whereas that is not necessarily the case for machines.

Finally, the cellular infrastructure – in particular the location and the settings of the base stations – is customized around the expected human position. This is why, for example, base stations are closer to each others in crowded urban areas and further apart in rural areas where the presence of humans is not very prominent. For all these reasons, M2M traffic needs to be thoroughly studied and the suitability of LTE, as well as other candidate solutions, needs to be assessed through customized performance evaluations.

LTE is basically composed of the Evolved Universal Terrestrial Radio Access Network (E-UTRAN) and the Evolved Packet Core (EPC). The E-UTRAN is supposed to be the segment that will be most affected by the introduction of massive M2M traffic. The resource scheduling in the uplink is based on an initial random access with a limited number of frequency slots to be used: this capacity might be the bottleneck when a large number of concurrent devices try to access the network. It is therefore important to assess whether the E-UTRAN is able to manage the traffic generated by M2M applications.

To examine this problem, we propose a web-based framework that can be used to characterize M2M traffic and study the suitability of standard LTE access mechanisms for massive M2M communications. The framework takes into account real geographic data on the position of machines, retrieved from public online databases. The traffic generated by realistic M2M applications is represented by means of three types of mathematical distributions (e.g., Poisson, Beta, and Deterministic). A network simulator was also implemented using Java to model the LTE random access procedure evaluating its performance. The proposed framework permits to visually analyze the performance of the LTE infrastructure, when simultaneously supporting multiple realistic M2M applications.

The remainder of this document is structured as follows: Section 2 contains a brief overview of the literature about M2M traffic characterization and performance analysis over LTE; Section 3 describes the implemented web application; Section 4 presents a case study with some numerical results; Section 5 contains the conclusions of this work.