DisGB Simulation

This repository contains the source code for the simulation tool from the DisGB paper (see below). The simulation tool is used to simulate inter-broker messaging for multiple rendezvous-based distributed pub/sub strategies. For a visualization of the broker setup from the paper, visit https://moewex.github.io/DisGB-Simulation/. To better understand how brokers exchange messages for each strategy, see Message Flows.

If you use this software in a publication, please cite it as:

Text

Jonathan Hasenburg, David Bermbach. DisGB: Using Geo-Context Information for Efficient Routing in Geo-Distributed Pub/Sub Systems. In: Proceedings of the 7th IEEE/ACM International Conference on Utility and Cloud Computing 2020 (UCC 2020). IEEE 2020.

BibTeX

@inproceedings{hasenburg_disgb:_2020,
	title = {{DisGB}: Using Geo-Context Information for Efficient Routing in Geo-Distributed Pub/Sub Systems},
	booktitle = {2020 {IEEE}/{ACM} International Conference on Utility and Cloud Computing},
	publisher = {{IEEE}},
	author = {Hasenburg, Jonathan and Bermbach, David},
	year = {2020}
}

How to build manually?

• Run mvn install in the root of the GeoBroker project, since it is a dependency.
• Run mvn package in the root directory of this project

How to run locally?

Preparation

• Install Java 11
• Build project or use pre-built Jar found at ./Broker-Simulation.jar
• Get the worldcities.csv data and store it at ./worldcities.csv.

To run the experiment:

mkdir tmp-results
java -jar -Xmx4G Broker-Simulation.jar -d tmp-results/ -i worldcities.csv -p exp --nBrokers 9 --nClients 1000 --nTopics 10 --pop 1000000 -e 900 --eGeofence 5.0 --sGeofence 10.0 --fieldSize 6 --history false

To learn more about command line argument options, run java -jar Broker-Simulation.jar -h.

How to visualize broker locations and strategies on a globe?

First, get results using the main Kotlin project. The results files need to be renamed for the visualization using the included script. Delete the sample files, copy all results into docs/CSV and execute ./docs/rename.sh. Now everything is in place. However, a web server is needed for the broswer visualization to work (security restrictions of browsers). To begin the visualization, follow these steps (if Python 3 is available):

• Navigate to docs
• Run python3 -m http.server 8080
• Open a web browser and navigate to http://localhost:8080/index.html.

Some notes:

• If you change the CSV file, beware browser caching! Changes might not display because of it.
• The loaded CSV file address is hardcoded. To change the name and port you have to change it in the code (when 8080 is unavailable).
• Also, depending on how many brokers the experiment includes, the visualization might not have a unique color for all of them. The colors repeat after they are exhausted. The color map includes 28 colors.

Further infos are inside of the README.md inside docs.

Simplifications:

• For all strategies, locations are flooded to all brokers. Thus, counting and comparing location updates does not make sense.
• While clients can be mobile, it is expected that they do not move out of the broker area of their original local broker.
• The unique identifier of a message is the origin clientId and tick