An Architecture and Prototype for Monitoring Distributed Simulations of Distributed Systems

Department of Computer Science and Engineering, Alma Mater Studiorum - Università di Bologna, Cesena - Italy

• Angelo Filaseta 0009-0004-6797-6814
• Danilo Pianini 0000-0002-8392-5409
• Angela Cortecchia 0009-0000-3650-7450

Requirements

This project requires a Java 17 or higher version. If you have no Java installed, it is recommended to install the latest LTS Java from AdoptOpenJDK (unless a new LTS is out and Gradle does not yet support it).

Experiments

Before proceeding with the experiments: - local ports from 1313 to 1412 should be available, along with the port 9090; - install the version of Python indicated in .python-version (or use pyenv); - Upgrade pip to the latest version: pip install --upgrade pip; - Install the requirements: pip install -r requirements.txt.

First of all, clone the repository and navigate to the project directory:

console git clone https://github.com/AngeloFilaseta/dsrt-2024-distributed-monitoring.git cd dsrt-2024-distributed-monitoring

Some Gradle tasks will be available for each experiment.

The experiments were originally executed on the following machine: - CPU: Intel(R) Core(TM) i7-8700T CPU @ 2.40GHz - RAM: 16GiB System memory - OS: Manjaro Linux

Viability

To execute the experiment highlighting the Viability goal, run the following command:

console ./gradlew runOneRootMonitorExperiment # UNIX & Bash Emulators gradlew.bat runOneRootMonitorExperiment # Windows

The default browser should open a new tab at http://localhost:9090/. If the browser does not open automatically, it is possible to manually open the page by visiting the URL. The Alchemist Monitor GUI should be visible now.

1) use the Selection Form to choose how the monitor configuration. 2) add the addresses of the clients to monitor using the Navbar. Three simulation are available to connect, at addresses:http://localhost:1313/, http://localhost:1314/, http://localhost:1315/, The Add All Available Clients client button will automatically add all the clients in one shot.

Press the Play button to start the simulations. Four plot will appear on the screen. Three of them, characterized by a green line, display the evolution of the local success metric for each client. The last one, characterized by a purple line, displays the same metric, aggregated over all clients.

Scalability

To execute the experiment highlighting the Scalability goal, run the following command:

console ./gradlew runOneRootResponseSizeExperiment # UNIX & Bash Emulators gradlew.bat runOneRootResponseSizeExperiment # Windows

The required time for completion was estimated to be around 1 hour and 30 minutes on the machine described in Experiments.

Raw metrics will be available in the evaluation/oneRootResponseSize directory. It is possible to generate the plot by running the following command: console python evaluation/scalability.py evaluation/oneRootResponseSize

If you want to re-generate the plots using the available data, run the following command instead:

console python evaluation/scalability.py evaluation/oneRootResponseSize_presaved

Efficiency

To execute the experiment highlighting the Scalability goal, run the following command:

console ./gradlew runOneRootLostUpdateUselessPollingExperiment # UNIX & Bash Emulators gradlew.bat runOneRootLostUpdateUselessPollingExperiment # Windows

WARNING: This experiments takes a long time to complete.

The required time for completion was estimated to be around 1 hour and 30 minutes on the machine described in Experiments.

Raw metrics will be available in the evaluation/oneRootLostUpdateUselessPolling directory. It is possible to generate the plot by running the following command: console python evaluation/efficiency.py evaluation/oneRootLostUpdateUselessPolling

If you want to re-generate the plots using the available data, run the following command instead:

console python evaluation/efficiency.py evaluation/oneRootLostUpdateUselessPolling_presaved