Part B: Utilization

Further questions were added to the questionnaire already discussed in Part A.  These were designed to discover more about how users make use of the Internet café and what their priorities are when they are choosing which Internet café to use.  

In addition, data was collected to show how much of the available computer and Internet bandwidth capacity were being used at different times.  These were directly measured using software.

Problems that occurred

Looking at a few responses from question 11, it could be seen that the question confused some respondents, or that they could not  answer it properly.  A number of respondents seemed to have given the same answer to all options, or answered the question partially. It became clear that this data might not be reliable.  Alternative methods of collecting the information were researched and considered as it was felt that it was important to know what users were actually doing with the connections they had access to.  A decision was made to use monitoring software to collect usage data. 

Usage monitoring

After some Internet research, a tool called Track4Win was discovered.  This could objectively measure what users do with the computers in the Internet cafés, in great detail.

Track4Win is installed in two parts: 

• A client program is installed on each computer to be monitored.  It records which applications are loaded and which are in focus at all times.  When accessing the Internet, it also records which URL's are visited and the caption for each page.

• A server program collects the usage data from each monitored computer and stores it in an Microsoft Access file.  The program has analysis and charting capabilities which allow the usage data to be interrogated, filtered by various criteria and reported.

A free demonstration copy of Track4Win allowed the client software to be loaded on five computers and provided full data collection facilities for up to fifteen days.  This was adequate for the purposes of this research. More detail about the monitoring software and sample outputs are shown in Appendix C.

The software was installed at the bigger of Supernet’s two Internet cafés, on five of the eighteen computers at that location.  For ethical reasons, a technician installed the client software on five random computers, and no information was collected on the identity of users.  This additional data collection was completed after the author had left Aceh, and the resultant data file was sent by email.  Monitoring occurred from May 27th to June 6th, 2006.  This produced nearly 50,000 detailed individual transaction records.  

Bandwidth consumption

The bandwidth provided via the VSAT to Supernet is 2Mbps downlink and 512kbps uplink contended at 10:1.  This connection has an additional 60kbps

of Dynamic Committed Information Rate (CIR) bandwidth.  For an explanation of contention and CIR, see the Appendix I.

The original business model called for about half of the available bandwidth to be resold to other businesses, and for the remainder to be used by Supernet's own Internet cafés and its associate charity organization.  The thinking behind this is that the other businesses are likely to have their peak usage during business hours, when the Internet cafés are less busy, and the Internet café peak usage is likely to be during the evening and night.  Since the business pays for the same amount of bandwidth 24 hours a day, it seems wise to try and generate revenue from it at all times too.  Indeed, the utilization of the bandwidth and optimization of revenue from it is thought to be as critical to the success of Supernet as the filling of aircraft seats is within an airline.

In order to measure the actual bandwidth consumed over a period, the monitoring capability built into a bandwidth management router from Mikrotik was used.  This recorded the download and upload bandwidth consumption at five minute intervals for the entire network (resold and Internet café bandwidth.) Data was also collected separately for different sub-nets representing each Internet café and each customer site using re-sold bandwidth.  More information about the Mikrotik monitoring capability is given in Appendix C.

Collation and analysis of results

The three sources of utilization data, from the survey, the usage monitoring and bandwidth consumption monitoring were treated separately, but together provided a full and detailed view of how the Internet café was being used during the studied periods.  

The survey data

As with the other results from the demographic survey, these results were collated in a spreadsheet and the resultant information was charted.

Usage monitoring data

The data from Track4Win proved a rich source of information.  The 50,000 transaction records could be analyzed several different ways.  One of the questions that needed to be answered was this: What proportion of the time are the computers in use during a 24 hour period?  Since users are charged by the hour, this information shows strong correlation with revenue. The proportion of time that the computers were in "active use" was recorded by Track4Win.  Although it is perfectly possible that a user is paying for time, but not actually using the computer, it was reasonable to assume that if keyboard and mouse are being manipulated with reasonable frequency, the computer is being used.  The accumulated "active use" in seconds per hour compared with the total number of seconds per hour produced a utilization percentage.  These numbers were graphed.  Utilization per day and for the whole monitored period were also calculated.

Next, it was interesting to discover what proportion of the time different applications are being used.  Different types of usage have different levels of bandwidth consumption.  Therefore, if there is a lot of bandwidth intensive use, the number of simultaneous users that can be supported by a given bandwidth is reduced. For example, if someone is using Yahoo‘s Instant Messaging client, they are likely to be using much less bandwidth than someone who is downloading a streaming video. The Track4Win data also showed every program and process running on the monitored computers, providing a very extensive list.  For simplicity, the author chose to focus on applications which were running actively (in focus applications rather than background applications) for more than 1% of the total time.  These were then categorized by application type.  For example, MS Internet Explorer and Mozilla/Firefox were both categorized as browsers.  This allowed a high level view of the main uses of the Internet to be generated.

Additionally, it was thought useful to find out the proportion of time users spend visiting different categories of Internet sites.  Question 11 on the survey was intended to obtain data on where users go when they are on the Internet.  As discussed earlier,  this data from the survey was not thought to be reliable.  In order to try and get the same data from the usage monitoring, a category field was added to the Access database produced by the Track4Win application, and each of the approximately 50,000 records was reviewed manually.  Looking at both the URL and the caption captured from the sites visited allowed most of the records to be allocated a letter matching the categories used in question 11.  It was decided to add two further categories letters: X for pornography and Z for system maintenance.  The reason for recording the visits to pornographic sites was two-fold.

• Firstly, visiting pornography sites is a subject that causes cultural resistance to the adoption of the Internet in some societies.  

• Secondly, pornography is suspected to have an impact on systems performance and maintenance: it usually contains a high proportion of graphical and video content which increases the average usage of bandwidth per user on the network.  It is also suspected as a major source of malware, causing further technical and  management issues.

It is recognized that this method of categorization of sites, from URL and caption is relatively crude.  Not having the time or the inclination to visit every site meant that it was not easy to judge quickly in some cases.  For this reason, a "not classified" category was also added.  Even so, it took approximately 40 hours to complete the manual categorization of these records.  The "active time" recorded against each activity was then presented as a pie chart.

Bandwidth consumption data

The Mikrotik router which was installed on the network to control bandwidth allocation to each customer also had the capability to record statistics about bandwidth consumption.  This produced ready-made graphs showing how much of the available bandwidth was being consumed.  This information indicates how much capacity is left for additional users, or for more speed to be made available to existing users.  The maximum and average throughput on each of the five days monitored was recorded and plotted with the contractual minimum bandwidth the satellite ISP is committed to provide. 

References:

- no references on this page.