As a follow up to our call for participation in the Internet measurement awareness survey, we are now presenting the outcome based on the responses during the exercise. The survey was an exciting activity and some of the results from it are quite interesting and unexpected, such as the finding that research is the third reason why our respondents are conducting measurement on the Internet. This study sheds new light by focusing on the users’ understanding and awareness of the measurement infrastructure, as well as highlighting the presence of a measurement infrastructure that could be potentially deployed within corporate networks and largely hidden from the research community.

We have results that tell us about the purpose of the different business categories for running measurements and why short-term simple measurement activities tend to be more popular than long-term measurement campaigns. The Internet measurement platforms that enable measurement on a large scale were also investigated and the results presented here too. It turns out that there is a need for more of these platforms to be deployed in the African region and there is also a great need to keep the platforms online at all times once deployed. An active training campaign is needed to build up the skills necessary to run a successful Internet measurement campaign, which we found to be lacking in the African region.

Distribution of Responses

We received 123 responses from 34 countries in Africa. These responses came from a number of business categories, which we grouped into five as follows (see the table below for details): Internet Service Providers (ISPs),  Academic Networks,  Internet eXchange Points (IXPs), Civil Society, Regulator/Government Agency, Enduser (Home/Mobile Broadband), and the ’Others’ category. The bulk of our responses, as shown in the table, came from the ISPs and it is not surprising as the major provider of Internet services and the group with the highest number of AFRINIC registered members. The second highest number of responses came from the academic networks group, followed by the 'Other' category. The spread of the responses across different countries and from different business interests and end users give some confidence that the survey data is representative of what could be obtainable in the region.

Business Category	Number of Responses
ISP (also comprised of telecom operators and wireless network providers)	35
Academic Network (Academic Institutions' networks and National Research and Education Networks)	34
Others (comprised of Enterprise Network, Community Network, ccTLD/DNS Operator, Data Centre, and Cable Operator)	22
End User	11
IXP	8
Civil Society	8
Regulator/Government Agency	5

  

Why Do Businesses Run Measurement?

The general idea about running measurement is to have a better understanding of a network. However, different businesses could be interested in different aspects of a network's behaviour or could have peculiar reasons for conducting measurement.  However, before finding out what their motivations are; do businesses in the African region really care about Internet measurement? It turned out, to some extent, they do; depending on what the measurement activity entails. Where the activity is of some significance and likely involving some specialised devices and the need to run for a good number of weeks/months, only 25.2% of the respondents/organisations said they were involved. This is in stark contrast to the 55.28% who reported to be conducting short-term, simple network measurement exercises apparently to understand what was happening on the network at that moment, which could be adequate in some instances. Moreover, the long term measurement campaigns are needed to have a bigger picture of the overall network performance.

As for the reasons for running measurements, network monitoring is the first choice (see the figure above), according to 67% of respondents. This is followed by the need to measure the quality of service (QoS) and the quality of experience (QoE) with 65% – respondents were allowed to choose more than one option, hence multiple selections were recorded for the listed purposes. With ISPs being one of the two dominant sources of the survey data, we could understand why network monitoring and QoS/QoE were prevalent. An ISP needs to constantly monitor the performance of its network and also have a clear understanding of the quality of the Internet connectivity that their clients are receiving. About 45% of respondents chose research as the reason for measurement and these were mostly from the academic networks; the group that produced the second largest number of respondents. Respondents from other categories also chose research as the reason for conducting measurement - up to 14% of ISP respondents, for instance, chose the research option. When we asked about the metrics of interest to get responses at a finer granularity, we received answers that showed diverse and different interests depending on the business entity. Moreover, bandwidth/throughput, latency, route reachability, and network utilisation are the most popular metrics in the order outlined here.

 

Measurement Infrastructure

For a regular, precise, and longitudinal measurement of the Internet in a particular region, infrastructures that are dedicated to periodically running Internet performance and topology measurements need to be deployed in that region. These infrastructures, popularly known as Internet measurement platforms, need to be hosted and kept online by networks in the region to create vantage points for collecting measurement data by interested parties on the Internet. About 42% of the respondents on our survey said they use a specific software on their PC to carry out Internet measurement. Applications mentioned include Uptime Robot, Netflow, View response PRTG, MRTG, Cacti, WANkiller, etc. From the established Internet measurement platforms, on the other hand, RIPE Atlas was the most popular platform, chosen by 34%, followed by Speedchecker with 21%. Respondents, to a lesser extent, also used other platforms including PerfSONAR, CAIDA Ark, M-Lab/NDT, Bismark Node, etc. Although most of the Internet access in Africa is via wireless networks, the use of mobile measurement apps is not as popular.

There is some level of disinterest in hosting these measurement platforms as 53% of our respondents said they do not host any measurement platform. This percentage includes respondents who used these platforms to carry out measurements as captured in the preceding paragraph. The other 47% that host – or hosted in the past – some measurement platform were mostly in the ISPs and academic network groups. No hosting was recorded with regulators/government agencies according to the survey data. RIPE Atlas probe comes out on top of the types of platforms hosted with Ookla speedtest server coming in as a distant second. As shown in the Likert scale below, about 70% of our respondents are either OK, Satisfied or highly satisfied with what the measurement platforms in the region currently provide. The question asked about their satisfaction level with the current measurement platforms not the platforms' availability in itself. The availability of measurement platforms in the region is abysmally low.

We also asked respondents about the tools they use to analyse/visualise the data collected and whether they are aware of any privacy laws regulating data collection in their countries. As for the analysis/visualisation, MS Excel (one of the four choices provided in the question, the others being Python, R, and 'Other') was the most popular used tool for analysis followed by Python then R. Respondents also mentioned the use of other tools such as tcpdump, netflow, ELK, etc, as part of the further information requested when respondents chose the 'Other' option. Respondents mostly chose more than one tool in their answers, which indicates that a combination of tools is utilised to analyse collected data. As for the laws governing data collection, the responses indicated that users were not aware of the existence of such laws in their countries.

 

Where We Are

At the moment, it is clear that Africa lags behind in generating high fidelity data regarding the continent's Internet operations. This is due to the fact that Internet measurement is yet to have a stronghold with the Internet community in the region – only 25.20% acknowledged to have run some measurement campaign in the past as reported earlier. Apart from the poor distribution of measurement platforms in Africa, another major problem is keeping the deployed devices up and running at all time, which is a key factor in increasing the number of measurement vantage points. We could see from RIPE Atlas records that of the 1026 probes delivered to users across the African region between 13 June 2014 and 10 May 2019, only 231 (22.5%) are currently connected, as depicted in the figure below.

We could also see from the survey responses that there is a great lack of awareness regarding the general area of Internet measurement. Furthermore, there is a lack of the skills necessary to run a successful measurement campaign. When we asked respondents for the reasons why the existing measurement platforms in the region were not well-utilised, 'Lack of technical know-how' was the most frequent answer.  Similarly, when we asked an open-ended question on what the respondents think is currently missing in the available measurement platforms/tools in Africa, we got 47 answers mostly pointing to the need to develop skills. 'Lack of well-trained personnel' was also the number one answer when we asked about the factors hindering Internet measurement in Africa.

Lastly, while privacy and security concerns could have an impact on the adoption of Internet measurement in a region, there is no evidence from our survey data to show that privacy and security laws have any bearing on the prevalence of measurement in Africa. We found that majority of the respondents were not aware of the existence of such laws in their countries.

 

Moving Forward

There has been a growing interest in measuring the different aspects of Internet connectivity and performance in Africa. This could be seen in the rise in measurement probe deployments across the continent and the increasing number of research literature related to the region that was produced in recent years. There are also issues, as highlighted above, bedevilling the Internet measurement activity in Africa. The following steps could help in ameliorating these issues discovered from the survey responses.

1. Increasing sensitisation on the subject by Internet bodies and groups in Africa, such as AFRINIC, *NOGs, ISOC, IGF, etc

2. increase the number of skills acquisition workshops where network administrators and users are taught the skills needed to conduct Internet measurement and data analysis.

3. There is a need for businesses, institutions, and individuals in Africa to host more probes/servers to increase the number of measurement vantage points in the region. It is vital that these devices are kept constantly up and running on the Internet at all times.

4. There is a need for research collaborations between researchers in different research institutions in the region and beyond and also collaborations with network engineers running the region's Internet.

5. African governments and regional bodies need to come up with laws and guidelines that will govern the usage of user traffic for the purpose of conducting Internet measurement.
6. Existing measurement platforms need to include features that are vital for Africa's peculiar set of issues, such as incorporating spectrum sensing and QoE metrics.
7. There is a need for different measurement platforms to interoperate and to also incorporate features that could be used to correlate different metrics. For example, congestion and QoE.