TopoConfluence is a synthetic simulation platform with various topologies and configurations to verify our fCoder design. It uses the topologies from Internet Topology Zoo (ITZ) [1] and Brite Topology Generator [2], configures the network (FNSS [3] for ITZ) and then use ns-3 [4] to simulate. After simulation, statistics like RTT are collected and bottleneck link of each flow is recorded for later usage (in our case, train a co-bottleneck detector).
Install python3 and pip3 by (Ubuntu assumed)
sudo apt-get install python3 python3-pip
Required python3 packeges include FNSS, networkx, numpy, xmltodict, etc. Install them by
pip3 install -r requirements.txt
No explicit installation of Brite or ns3 is required if you run simTop.py first.
Specifically, it will help you download ITZ topology(if there isn't any), make Brite(without --run-only), and configure ns-3(if not before and without --dry-run).
You can run python3 simTop.py at the beginning to get these all done.
The main procedures:
- Download the ITZ dataset and randomly sample some topologies (in
simTop.py); - Parse the topologies and configure the networks (delay, backbone bandwidth) by FNSS parser (in
src/topoSurfer.py); - Configure ns3 (in
simTop.py); - Run simulations with choice of flow settings and leaf bandwidth (in
src/confluentSim.py).
As a user, the only thing we need to do is to use simTop.py with right arguments. The following way is tested.
First, Generate 64 topologies randomly without running simulation over them:
python3 simTop.py -s 64 --dry-run
After this step, 64 topologies are collected and parsed to XML files stored in TopoSurfer/xml for later use.
Second, Run over a specific topology (i.e. 27th XML file) with flow number ranging from 10 to 16 using 64 processes
python3 simTop.py -c 64 -r 10:16 --run-only -x 27
It will use choose 1 from existing XML files and run 7 * 9 = 63 runs on 64 cores
(default 9 different settings for each flow number, see src/confluentSim.py for more detail).
Note that if you don't specify --run-only when there are existing topologies in TopoSurfer/xml, the script will again add some random topologies, increasing the total number of topologies.
At last, data will be collected in Flows_MIN:MAX_XML-NAME/.
dat/ stores statistics per flow in files like DATA_MID_FLOW-NUM.dat.
For instance, RttLlr_5847_0.dat means Rtt and Llr data of flow 0 in run 5847. The internal row format of files except bottleneck is
Time Data [Data2 if exists]
Row format of bottleneck_MID.txt is as follows:
Flow Bottleneck_link # 4294967295 or -1 for no bottleneck
info/ stores flow settings for each run in flow_info_MID.txt. The internal row format is
Source Destination Leaf_bandwidth # target flows
Source Destination # cross traffic
log/ stores the ns3 logs, which are useful for debugging.
Our simulation with Brite topology mainly focuses on the dumbbell structure between 2 ASes, in which case we can control the co-bottleneck we want better. Since ns3 integrated Brite module, we don't need a parser like FNSS to help us parse the topology. So the main procedures:
- Configure the network (inter-AS bandwidth) by generating the Brite configure file;
- Scan parameters (edge bandwidth, cross traffic) and run ns3 simulations.
See our paper for more detail about the network settings.
As a user, to simulate a network with 7 target flows with random seed 20 using 24 processes:
python3 simBrite.py -c 24 -r 7:7 -b 20
[1] Knight, Simon, et al. "The internet topology zoo." IEEE Journal on Selected Areas in Communications 29.9 (2011): 1765-1775.
[2] Medina, Alberto, Ibrahim Matta, and John Byers. "BRITE: a flexible generator of Internet topologies." (2000).
[3] L. Saino, C. Cocora, G. Pavlou, A Toolchain for Simplifying Network Simulation Setup, in Proceedings of the 6th International ICST Conference on Simulation Tools and Techniques (SIMUTOOLS '13), Cannes, France, March 2013
[4] Riley, George F., and Thomas R. Henderson. "The ns-3 network simulator." Modeling and tools for network simulation. Springer, Berlin, Heidelberg, 2010. 15-34.