^{1}

^{*}

^{2}

^{2}

^{1}

^{3}

We present a problem for benchmarking the robustness of cellular up-links, in an automatic weather station (AWS) testbed. Based on the problem, we conduct a small-scale measurement study of robustness, where the AWS is equipped with four (4) cellular modems for weather data delivery. The effectiveness of up-links is challenging because of overlapping spatial-temporal factors such as the presence of good reflectors that lead to multi-path effects, interference, network load or other reasons. We argue that, there is a strong need for independent assessments of their robustness, to perform end-to-end network measurement. However, it is yet difficult to go from a particular measurement to an assessment of the entire network. We extensively measure the variability of Radio Signal Strength (RSSI) as link metric on the cellular modems. The RSSI is one of the important link metrics that can determine the robustness of received RF signals, and explore how they differed from one another at a particular location and instant time. We also apply the statistical analysis that quantifies the level of stability by considering the robustness, referring short-term variation, and determines good up-link in comparison to weak one. The results show that the robustness of cellular up-links exists for an unpredictable period of time and lower than one could hope. More than 50% of up-links are intermittent. Therefore, we plan to extend our work by exploring RSSI thresholds, to develop a classification scheme supporting a decision whether a link is either intermittent or not. This will help in normalizing the level of stability, to design the RSSI estimation metric for the robust routing protocol in weather data networks.

Terrestrial wireless communication networks encompass the use of cellular networks or dedicated wireless networks to pass information between terminals. For instance, AWS networks [

Up-link refers to a one-way interconnecting between a modem and base station for the purpose of transmitting weather data to the central repository. Cellular up-link is dynamic, with the quality of which can change considerably with time and distance, due to the presence of various factors such as the presence of good reflectors (e.g. metal, walls, woods and glass) that lead to multi-path (reflections, fading, diffraction) effects, interference, network load or other reasons. Hardware miss-calibration and antenna position are other two additional factors that may affect signal propagation [

Intermittent links sometimes exist as perfect links and sometimes don’t exist for an unpredictable period of time. When the topology of the network is at least predictable, then the behaviors of intermittent links can be analyzed. So, in order to understand the robustness of up-link, is necessary to exploit its stability, and to provide meaningful information about its surrounding. The study of robustness of up-links [

This paper, attempts to answer the following key question, “whether RSSI is suitable metric to determine what level of robustness is expected?” We aim to benchmark the variability of RSSI on cellular up-links, to determine its robustness at a particular location and instant of time. In order to tackle this objective, the paper investigates the spatial and temporal effects on the RSSI, a measure of the radio frequency (RF) signals a power level that a node is receiving. Finally, we focus on understanding what robustness eventually reaches, a steady minimum value by considering the level of stability, whether is good, intermediate or bad. Thus, our paper consolidates the significance of considering RSSI as a link metric, and explores its accuracy to estimate the robustness against spatial-temporal factors, referring short-term variation. This will help in normalizing the level of robustness, to design RSSI estimation metric for the robust up-link routing protocol in weather data networks. The stability has to avoid long-term variation so that the routing does not have to reconsider alternative up-link, which is energy consuming.

The paper is organized as follows. Section II describes a typical scenario of other related works. Section III presents the experimental setup. Section IV describes some experimental results and analysis. Section V draws conclusions and future work.

In this section, we review related work, including efforts made and challenges faced. In addition, we highlight the gaps that have influenced our work. The works in [

There are several methodologies [

Other studies of the robustness of networks, have mainly focused on describing how a given performance metric of the network is affected when terminals or nodes are removed [

The experimental setup is based on AWS testbed, equipped with wireless sensor network (WSN). It is a networked system of interconnected nodes communicating wirelessly and reporting their measurement data to a central repository. In more detail, each such sensor network node has typically several parts: a radio transceiver with an internal antenna, a microcontroller (MCU) an Atmel ATMega128RFA1 [

In

The AWS unit is powered from at least two different sources: the grid-power and an embedded form of energy harvesting (e.g. solar panels). The gateway and sink node are powered by grid source, the remaining nodes are powered by a 15 W solar panel and ultra-capacitor consisting of two Lithium Ion Capacitor (LIC) via LM2596S step-down converters (regulators). The PSU regulator LM2596S converts any input voltage between 4 - 24 V to a constant voltage (3.6 V) required to operate.

The experimental environment is on the 2^{nd} floor in a four (4) floors building, consisting of various rooms. The environmental setup is a high ceilinged building, constructed from concrete blocks (granite). It has thick concrete walls (0.25 m thick outer walls, and 0.15 m thick walls between rooms), 5 m high ceilings, heavy wooden doors, tables and other office equipment. During measurements, the gateway is positioned at three different locations (L1, L2, L3). The distance between L1 - L2 is about 4 m, the distance between L2 - L3 is about 3 m and the distance between L1 - L3 is about 6 m. Owing to heavy attenuation by walls and doors, link outage in the environment is possible large.

The RSSI measurements are conducted by connecting four (4) cellular modems (SIM800L, HSDPA HUAWEI E153, HSDPA HUAWEI E173 and RoHS HSPA ZTE MF665C) at the gateway via standard USB hub 2.0. The modems make connections to the INTERNET from different mobile operators including Halotel, Tigo, Airtel, and Vodacom. During measurements, the experiment is repeated at each location, and RSSI is recorded in real time. The experiment is conducted with an average time of one hour, and the total amount of time it took for a modem to respond to a request for RSSI value is a second (it is the sum of service time and wait time). An open question “How many readings are needed for analyzing a robust link?”, remains unanswered. However, various studies claimed to attain a reliable analysis of large historical data above 50 packets [

When benchmarking the robustness of cellular up-link, we need to find ways to measure the network performance and determine when links are intermittent, as each network is different in nature and design. Whatever the approach we take to the problem, we use RSSI as link metric that reflects the performance status of the up-link. Therefore, in this section, we present the results regarding our measurements, and analyze the variability of RSSI, and discuss the statistical properties of the measured data, and finally benchmark the level of stability by considering the robustness of up-link, referring short-term variation.

The RSSI varies as a function of time and distance. To analyze the variability of RSSI, we consider spatial and temporal variations between receiver and transmitter. Several statistical numerical measures for describing the variation, or spread, or dispersion, of RSSI, are considered including, minimum, maximum, median, range, mean, variance, standard deviation, and coefficient of variation.

Spatial variation: RSSI values vary as a function of a distance between transmitter and receiver. In free space, RSSI is inversely proportional to the squared distance between the transmitter and the receiver.

P r ( d ) = C f ( P t / d 2 ) (1)

where P r is the received power, C f constant depending on a transceiver, P t transmitting power and d distance.

Tables 1-3 indicate the statistical data for the consistent four (4) mobile operators at locations L1, L2, and L3. For instance, the range difference at two locations L1 and L2 is 2 dBm for Halotel, 32 dBm for Tigo, −4 dBm for Vodacom and 11.3 dBm for Airtel. In the second experiment, the range difference between two locations L1 and L2 is 23 dBm for Halotel, 6 dBm for Tigo, −2 dBm for Vodacom and 5.7 dBm for Airtel. This means, ranges of RSSI are not constant, but this does not tell us how the RSSI observations are distributed between the smallest and the largest ones. The only information we have from the range is the distance between the smallest and the largest measurements. So, the spatial distribution of base stations is not uniform, since the capacity of connections in an area is basic design criteria for mobile networks, that the capacity of the base stations in any given area at a certain point in time reflects the user mobility, and depends not only on the location of the base stations but also the availability of frequencies that can change over time to follow the users. Since we are interested in the variability of the RSSI, in other words, we estimate the level of stability by calculating the RSSI mean.

For instance, from the first measurement, it is seen that the RSSI mean variations of 11.8 dBm for Halotel, −25.9 dBm for Tigo, 7 dBm for Vodacom and −2.3

Exp. | Operator | Min | Max | Median | Range | Mean | Variance | SD | CV |
---|---|---|---|---|---|---|---|---|---|

1. | Halotel | −59.0 | −51.0 | −51 | −8 | −51.9 | 0.387 | 0.622070 | 2.03 |

Tigo | −79.2 | −71.2 | −77 | −8 | −76.2 | 2.279 | 1.509668 | 8.2 | |

Vodacom | −65.0 | −51.0 | −61 | −14 | −60.1 | 1.569 | 1.252845 | 4.74 | |

Airtel | −65.0 | −58.4 | −63 | −6 | −62.7 | 0.487 | 0.697729 | 2.78 | |

2. | Halotel | −59.0 | −53.0 | −57 | −6 | −56.2 | 0.4328 | 0.657856 | 2.32 |

Tigo | −69.2 | −69.2 | −69 | 0 | −69.2 | 0 | 0 | 0 | |

Vodacom | −63.0 | −51.0 | −61 | −12 | −60.9 | 0.1333 | 0.365161 | 1.40 | |

Airtel | −63.7 | −58.4 | −62 | −6 | −62.1 | 0.0619 | 0.248838 | 0.98 | |

a. Statistical properties of RSSI at location L1.

Exp. | Operator | Min | Max | Median | Range | Mean | Variance | SD | CV |
---|---|---|---|---|---|---|---|---|---|

1. | Halotel | −69.0 | −59.7 | 25 | −10 | −63.7 | 1.1323 | 1.064108 | 4.32 |

Tigo | −89.2 | −49.2 | 31.9 | −40 | −50.3 | 10.926 | 3.30551 | 10.55 | |

Vodacom | −73 | −63.1 | 23 | −10 | −67.1 | 1.2189 | 1.104055 | 4.81 | |

Airtel | −75 | −57.7 | 26.3 | −17.3 | −60.4 | 1.4658 | 1.210702 | 4.60 | |

2. | Halotel | −82.0 | −53.0 | 29 | −29 | −56.2 | 0.8933 | 0.945119 | 3.33 |

Tigo | −91.2 | −85.2 | 12.9 | −6 | −87.2 | 0.6317 | 0.794781 | 6.15 | |

Vodacom | −73.0 | −63.0 | 23 | −10 | −67.0 | 1.1985 | 1.094746 | 4.76 | |

Airtel | −79.4 | −67.7 | 21.3 | −11.7 | −70.1 | 0.3559 | 0.596536 | 2.78 | |

b. Statistical properties of RSSI at location L2.

Exp. | Operator | Min | Max | Median | Range | Mean | Variance | SD | CV |
---|---|---|---|---|---|---|---|---|---|

1. | Halotel | −65.0 | −57.0 | 26 | −8 | −60.9 | 0.3871 | 0.622205 | 2.39 |

Tigo | −91.2 | −87.2 | 10.9 | −4 | −90.3 | 0.6963 | 0.834423 | 7.34 | |

Vodacom | −73.0 | −65.0 | 22 | −8 | −68.8 | 1.7010 | 1.304225 | 5.90 | |

Airtel | −75.7 | −70.4 | 22 | −5.3 | −73.3 | 0.3251 | 0.570218 | 2.58 | |

2. | Halotel | −69.0 | −57.0 | 26 | −12 | −61.3 | 1.2669 | 1.125559 | 4.35 |

Tigo | −103.0 | −55.7 | 10.9 | −47.8 | −79.9 | 66.545 | 8.16365 | 49.39 | |

Vodacom | −75.0 | −63.0 | 24 | −12.0 | −66.6 | 1.3819 | 1.17556 | 5.07 | |

Airtel | −77.7 | −60.4 | 20.63 | −17.3 | −69.3 | 6.1055 | 2.47093 | 11.3 | |

c. Statistical properties of RSSI at location L3.

dBm for Airtel are observed between locations L1 and L2. Thus, by comparing the results from different operators we conclude that, the location granularity of 3 m is a considerable difference in terms of RSSI range and mean. The up-link is stable when the RSSI mean is above −60 dBm. But, when the RSSI value is less than −65 dBm, it varies a lot. This means, the up-link is unpredictable if RSSI variability overlaps with a certain RSSI mean.

Temporal variation: We also observe that, the RSSI captured during the daytime may not be the same as the one captured during evening time, due to changes in the environmental characteristics. Several studies confirmed that the temporal variation of RSSI is due to the changes in the environmental characteristics, such as climate conditions (e.g. temperature, humidity), human presence, Interference (e.g. WiFi) and obstacles. Also, the variation of the RSSI may also be due to either constructive or destructive interference in the deployment environment.

Since wireless up-links are highly dynamic, with significant changes in time and surrounding activities. To account for these variations, we measure RSSI from different mobile operators, knowing the amount of standard deviation and measure its coefficient of variation. As shown in

We analyze the level of stability by considering the robustness of the up-links. Consider the following equation:

Robustness min = efficiency min / efficiency max (2)

The robustness value is normalized between 0 and 1, and it measures the relative loss of efficiency caused by the poor up-link. If the intermittent link does not impact the efficiency, then its robustness is 1, while if the intermittent link destroys the efficiency, then the robustness drop to 0. So, in order to do that, we first compute RTT when the packet size of 236 and 277 bytes are sent to a repository. The RTT quantifies network delay as the duration of time taken by a packet to reach the repository from the gateway plus the duration of time taken by a packet to reach the gateway from the repository. The experiment is repeated at different locations (e.g. L1, L2 and L3), and the RTT mean over all the experiments is obtained as a function of up-link efficiency. It is important to note that RTT measurements are biased by differences in the paths between different modems and the repository they communicate with. The delay between the gateway and the repository decoding it, is mainly a function of the packets travel time, and processing time at the gateway the information traverses.

Latency = time ack + 2 * time delay (3)

As expressed in the equation above, the RTT or latency is the time for the signal to propagate data or packet from a gateway to a central repository, and send back to a gateway. But the facts expressed by channel efficiency ratio, differ among networks because propagation delay is only one of the various factors affecting the robustness of cellular up-links. Therefore, our goal in this part is to minimize packet losses so that we can obtain the best possible case for robust up-links. An important aspect is to study the link efficiency, which is essential for computing the robustness of cellular networks. Therefore, channel efficiency determines a fraction of the transmission capacity of a particular up-link that contains the amount of data packet to be delivered to a repository. This measures the relative loss of efficiency caused by intermittent links. We expect that the maximum channel utilization can be determined, where data are delivered without transmission errors.

channel efficiency = time data / ( time data + time ack + 2 ∗ time delay ) (4)

In

As shown in

throughput. Another property of interest is the distribution to which the varying throughput belongs. The distributions at different time are the key properties for studying the robustness of intermittent links. This property is important since we are targeting a time for which throughput can be detected from RSSI. The observations confirmed that the RSSI is not proportion with throughput.

In this paper, we focus on the problem of benchmarking the robustness of cellular up-links based on spatial and temporal effects. We analyze the empirical data from AWS testbed with four (4) cellular up-links on the indoor environment. We show empirically that, the RSSI variations are quite significant. The variations are amplified even further due to some burst transmission, which exists for an unpredictable period of time. To better understand the robustness by considering the level of stability, we consider the following: 1) We present the variability of RSSI based on spatial and temporal effects such as the presence of good reflectors (e.g. metal, walls, woods, and glass) that lead to multi-path (reflections, fading, diffraction) effects, interference, network load or other reasons. 2) We present the statistical analysis of RSSI variation, which has distinct values that can discriminate the quality of links. The spatial variation results at least 11 dBm mean deviation on Halotel whereas location granularity of 3 m leads to up to 11 dBm mean variation. Further, the temporal variation can reach in between −7 dBm and 4 dBm with a time-scale variation about five (5) hours. 3) We present the variation of RSSI against latency, efficiency, throughput and evaluate the level of stability by considering the robustness of the up-link against failures.

Thus, we conclude that an up-link with the high level of robustness guarantees a successful packet reception, as the difference between the minimum and maximum efficiency decreases. Our experiments show that Halotel, Vodacom, Airtel and Tigo are robust by 16.7%, 10%, 7.14%, 6.25% respectively. For many up-links particular when the robustness is small, we can see all the possibilities of the intermittent link. Our future work, intends to extend to explore RSSI threshold, to develop a classification scheme in supporting a decision whether a link is either intermittent or not. This will help in normalizing the level of stability, to design the RSSI estimation metric over intermittent links, for the robust routing protocol in weather data networks.

The authors would like to acknowledge the financial support of NORAD in collaboration with Dares Salaam Institute of Technology for supporting this work under WIMEA-ICT project, with the objective to improve weather information management in East Africa.

Kondela, E.A., Nungu, A., Matiko, J.W., Otim, J.S. and Pehrson, B. (2018) Benchmarking the Robustness of Cellular Up-Links in Automatic Weather Station Networks. Communications and Network, 10, 78-92. https://doi.org/10.4236/cn.2018.103007