About

This repository contains the data traces and source codes for multi-path TCP (MPTCP) measurement on high-speed rails. It also contains the scripts for data analysis and plotting.

Please cite the paper as follows (or use this bibtex record).

• Li Li, Ke Xu, Tong Li, Kai Zheng, Chunyi Peng, Dan Wang, Xiangxiang Wang, Meng Shen, Rashid Mijumbi. A Measurement Study on Multi-path TCP with Multiple Cellular Carriers on High-speed Rails. Annual conference of the ACM Special Interest Group on Data Communication on the applications, technologies, architectures, and protocols for computer communication (ACM SIGCOMM), pp. 161-175, 2018.

Data Traces

data_traces includes the traces for elephant and mice flows collected.

Tools and Scripts

The directory structure in src is as follows:

HSR Measurement
|-- android            # Code of Mobinet for measurement
|   \-- ...
|-- bash            # Code for measurement
|   \-- ...
|-- c               # Code for create txt files
|   \-- ...
|-- matlab          # Code for evaluating the performance
|   \-- ...
\-- README.md

How to Use

Measurement

First of all, the Android source code of MobiNet output an APK, which can be installed on the test smartphones.

For measurement, you could copy the bash code into Linux environment and just run run.sh. Notice here, in our experiments, we requires subflows using different WLAN environment. That is, for each subflow, the corresponding WLAN should have a different SSID.

loop.sh is used for countint the bytes in/out from each port

monitor.sh is used for monitoring if MPTCP works well

Our experiments will collect data on both ends, for each experiment, data is stored in ./<HSR-ID>/<EXPR-ID>/ and organized as following:

.
|-- HSR-ID                          # HSR ID, like C2006
|   |-- EXPR-ID                     # Experiment ID, starts from 1
|   |   |-- ifaceX                  # Port name, like eth0, wlan1, etc.
|   |   |   |-- THP-File            # .thp file collected by loop.sh
|   |   |   |-- PCAP-FILE           # .pcap file collected by tcpdump
|   |   |   |-- carrier             # the carrier of ifaceX
|   |   |   |-- interface           # the name of ifaceX
|   |   |   \-- ip_address          # the IP address of ifaceX
|   |   |-- buffer_size             # experiment buffer size
|   |   |-- congestion_control      # experiment algorithm like reno, lia, etc.
|   |   |-- download_size           # download file size
|   |   |-- end_time                # experiment end time
|   |   |-- expr_type               # experiment type (Obsolete)
|   |   |-- scheduler               # MPTCP scheduler
|   |   \-- start_time              # experiment start time
|   \-- ...
\-- ...

We also record the basic experiment information in global.log file with following format:

EXPR-ID CARRIER-TYPE EXPR-TYPE CONG RMEM  WMEM  SCH     FILE    SV-IP   START-TIME END-TIME  
1       MOBILE       0         Reno 16384 16384 Default 1GB.dat 1.2.3.4 12345.123  12465.234 

(continue with the last line)

CL-PCAP-PATH               CL-THP-PATH               SV-PCAP-PATH         CARRIER FILE-SIZE
~/C2008/1/wlan2/wlan2.pcap ~/C2008/1/wlan2/wlan2.thp ~/C2008.1.wlan2.pcap Mobile  1GB

Analyzing

For analyzing the performance, first you should create txt file from pcap file with our tool pcap2txt. You can build and use pcap2txt as following:

g++ pcap2txt.cpp -o pcap2txt
find . -name "*.pcap" -exec ./pcap2txt {} {}.txt \;

After txt file processing, you can use Matlab to analyze its performance. Here we provide several kinds of performance evaluation, like throughput, RTT, bytes-in-flight, loss-packet-rate, packet retransmitting, and reordering. They are all included in Matalb directory. We also add a example.m to indicate how to use them.

Matlab
|-- cwnd.m              # Code for calculating bytes-in-flight
|-- disorder.m          # Code for calculating reordering
|-- example.m           # Example code for use
|-- retrans.m           # Code for calculating retransmitted packets
|-- rtt.m               # Code for calculating RTT
|-- throughput.m        # Code for calculating throughput
\-- txt2stream.m        # Read txt file into matrix format