• No results found

3 M ETHODOLOGY

3.5 Exploratory Steps of Analysis

3.5.6 Iteration 6: Autonomous System Relationships

The k-core decomposition of the Graph Visualisation Analysis in Iteration 5 indicated an important set of densely connecting Autonomous Systems for each of the three Tamil Nadu mobile broadband operators. Merely identifying these structural bottlenecks was not satisfactory. Hence, Iteration 6 aimed to reveal the economic nature of the most important mobile broadband operator networks’ relationships between the influential Autonomous Systems. The outcomes of Iteration 6 are reported in section 4.4 of Chapter 4.

Secondary Datasets

To reveal the economic nature of the Autonomous System relationships, we first opened the CAIDA (2016a) AS-Rank website and filtered the visualised AS-Rank dataset to fit the data-collection time of this dissertation’s time horizon (see section 3.3.4). The filtering was therefore set as ‘Dataset: 2015-02-01 IPv4’. Once the table view updated, it revealed 49,874 Autonomous Systems containing information on the Customer Cone Size (Number of Autonomous Systems and IPv4 prefixes), the percentage of the AS of all Autonomous Systems, IPv4 prefixes and the AS Transit Degree. Next, we sorted the table view by ‘number of ASes in customer cone’, resulting in another reload of the table view. We then downloaded the *.html table contents and placed them in an Excel file with the following steps:

• Right-click on the CAIDA (2016a) AS-Rank website > view page source • Copying all the content in the cache.

• Pasting the copied content in a new text document with ‘Ctrl+v’ • Saving the file as ‘as-rank.html’,

• Converting the ‘as-rank.html’ file into a *.csv file by using Conversiontools (2012).

• Saving the file as ‘CAIDA_AS_Rank_Data_01-02-2015.csv’ and ‘CAIDA_AS_ Rank_Data_01-02-2015.xlsx’.

Once downloaded, we manually searched the resulting file according to the most important information (Customer Cone Size, Number of IPv4 prefixes and Transit Degree) for those Autonomous Systems that the k-core decomposition in Iteration 5 revealed as most interesting to our case study. Additionally, we referred to the Border Gateway Protocol (BGP) Routing Tables of Hurricane Electric (2016). This helped us to provide a more thorough understanding of our operator networks, where applicable. Some of these tables are stated in the Appendices. Next, we downloaded the secondary CAIDA(2016b) AS-Relationship dataset by filling out the prompted user info request on the CAIDA website. This secondary dataset helped us to test our three Tamil Nadu mobile broadband operator networks through the economic nature of Autonomous System relationships, where a relationship could either be of peer-to-peer, customer-to-provider, or provider-to-customer nature. The downloaded file was saved as ‘CAIDA_AS_Relationship_Data.txt’.

Once obtained, we opened the operator’s Autonomous System edge tables that resulted through the Gephi (2016) export (containing the source and hop Autonomous System Numbers) in Iteration 4 and saved the file as ‘Aircel_ASRank_ASRel_mapping.xlsx’ and ‘Aircel_ASRank_ASRel_mapping.csv’, respectively. The data itself is stored in a sheet named e.g. ‘Bharti_Airtel_edges_after_Gephi’ for Bharti Airtel. We then added two new sheets to the file, named ‘AS_rel’ and ‘Transit_Table’. The first sheet, ‘AS_rel’, contained the AS-Relationship data from the ‘CAIDA_AS_ Relationship_Data.txt’, filtered by those Autonomous Systems of the respective mobile broadband operators. This data resulted from the secondary CAIDA (2016b) AS-Relationship dataset. The second sheet, ‘Transit_Table’, contained the Transit Degree obtained from the ‘CAIDA_AS_Rank_Data_01-02-2015.xlsx’ file from the secondary CAIDA (2016a) AS-Rank dataset.

Next, using Excel’s INDEX algorithm (see algorithm in Excel file), we fused the source and hop Autonomous System Numbers in the first sheet (e.g. ‘Bharti_Airtel_edges_after_Gephi’) with their associated Transit Degree from the elaborated ‘Transit_Table’ sheets, into two new columns in the first sheet, named ‘Source_Transit’ and ‘Target_Transit’. We then fused the source and hop Autonomous System Numbers, together with their corresponding ‘Source_Transit’ and ‘Target_Transit’, which resulted in two new columns containing the ‘SourceASN: TransitDegree’ and the ‘HopASN: TransitDegree’, respectively. To prepare for the later Gephi (2016) import, we named these new columns ‘Source’ and ‘Target’ and saved the file as ‘AS_Rank_Analysis.xlsx’.

Since the goal of this analysis is to explore the economic relationships of the Autonomous Systems in the operator networks, we next fused the file with the secondary CAIDA (2016b) AS-Relationships dataset. For this purpose, we first imported the downloaded ‘CAIDA_AS_Relationship_Data.txt’ into Excel, saved the sheet as ‘AS_rel’ and the file as ‘AS_Rel_Analysis.xlsx’ in the ‘../Step6/Secondary_CAIDA(2016b)_AS_Relationship /’ folder. The AS-Relationships in the ‘AS_rel’ sheet are represented by three columns, named ‘AS1’, ‘AS2’ and ‘rel’, indicating the relationship between two Autonomous Systems. Next, we created three sheets named ‘Aircel’, ‘Bharti Airtel’ and ‘Vodafone’ and imported the respective operator edge-tables from Iteration 4. Each of these three sheets contained only a ‘Source‘ and a ‘Target’ column. To uniquely match the relationships for the three operator networks, we first combined the ‘AS1’ and ‘AS2’

columns in the ‘AS_rel’ sheet into a new column, named ‘AS1AS2’. This column was useful for unique referencing purposes. Similarly, we fused the ‘Source’ and ‘Target’ columns for each of the three mobile broadband operator sheets (see above) into a column named ‘SourceTarge’ column. We then added a new column named ‘INDEXMATCH’ to each operator sheet, where we used a combination of Excel’s MATCH and INDEX algorithms, linking the ‘AS1AS2’ column of the ‘AS_rel’ sheet with the ‘SourceTarget’ columns of the operator sheets to reveal the corresponding ‘rel’ column of the ‘AS_rel’ sheet for each of the given operator sheets. The result represented the AS-Relationships (‘peer-to-peer’, ‘customer-to-provider’ and ‘provider-to-customer’) for the Autonomous Systems in our three operator networks, represented by a ‘0’ for a peer-to-peer relationship, a ‘-1’ for a provider-to-customer relationship and a ‘1’ for a customer-to-

provider relationship.

Next, looking at a combination of both, the ‘AS_Rel_Analysis.xlsx’ and the ‘AS_Rank_Analysis.xlsx’, we reported some preliminary findings in section 4.4 for each of the three mobile broadband operators. Next, we created a new file for each of our case studies’ three operator networks named e.g. ‘Aircel_ASRank_ASRel_ mapping_(Edges).xlsx’ for Aircel. In this file, we copied the ‘Source_Transit’ and ‘Target_Transit’ columns from the ‘AS_Rank_Analysis.xlsx’ file as well as the corresponding ‘rel’ column from the ‘AS_Rel_Analysis.xlsx’. This allowed us to measure the economic relationships per mobile broadband operator. For this purpose, we created a sheet named ‘analysis’ for each of the three operator files and counted the number of edge observations, the number of edge-weights and the percentage of edge weights of all edges per AS-Relationships, as stated above. Our findings were again reported for each mobile broadband operator. Moreover, to visualise the economic relationships between the Autonomous Systems in the three operator networks, we visualised the three networks again in a two-dimensional Euclidean space. For this purpose, we generated a *.csv file (named e.g. ‘Aircel_ASRank_ASRel_ mapping_(Edges).csv’ for Aircel) from the respective Excel files (named ‘Aircel_ASRank_ASRel__mapping_(Edges).xlsx’ for Aircel, for example) and imported the *.csv files again as edge table into Gephi (2016) and saved them (named ‘Vodafone_ASRank_ASRel_mapping.gephi’ for Vodafone, for example). The key here was to set the ‘rel’ column as relationship label when importing the dataset into Gephi (2016). This allowed us to colour the relationships, or edges, between each set of Autonomous Systems. Here, we coloured a peer-to-peer relationship between a set of

Autonomous Systems ‘green’, provider-to-customer relationships ‘red’, ‘customer-to-

provider’ ones ‘blue’ and ‘yellow’ for undetected ones. The resulting graph visualisations

were then saved, using weighted and non-weighted edges, as *.png files (named, for example, as ‘Aircel_Relationships_w.png’ for the Aircel graph visualisation). Next, we again reported our findings, for each of the three operator networks.