Opportunistic Network Monitoring

Research project on opportunistic network traffic monitoring.

For credits, please reference this publication:

S. Magnani, F. Risso and D. Siracusa, "A Control Plane Enabling Automated and Fully Adaptive Network Traffic Monitoring With eBPF," in IEEE Access, vol. 10, pp. 90778-90791, 2022, doi: 10.1109/ACCESS.2022.3202644.

Installation

This project requires DeChainy to be correctly set up in you system, either bare-metal or using containers. For performance reasons, we preferred the bare-metal installation, whoch can be installed by following this guide.

The following requirement is matplotlib for creating charts.

Concerning the eBPF programs, you must have a recent Linux kernel >= v5.6, since some tests use BPF_QUEUEs which have been recently introduced.

Finally, for the tests that require network traffic to be sent, the client requires a working build of MoonGen under the home directory of the user used for the tests. Please refer to their guide for installing and setting up the correct driver for your network interface.

Project Structure

.
├── adaptiveness
│   ├── ebpf.c
│   ├── __init__.py
│   └── __main__.py
├── erase
│   ├── ebpf.c
│   ├── __init__.py
│   └── __main__.py
├── nprobe
│   ├── ebpf.c
│   ├── __init__.py
│   └── __main__.py
└── swap
│   ├── ebpf.c
│   ├── __init__.py
│   └── __main__.py
├── moongen.lua
├── plotter.ipynb
└── set_irq_affinity.sh

• adaptiveness: Test case, containing an eBPF probe for measuring the impact of having a non-adaptive (nProbe), adaptive (pre-defined list of possibilities), and fully-adaptive solution
• erase: Test case, containing an eBPF probe for measuring the impact of some populate-erase operations on eBPF maps, including their most recent optimized versions (batch operations)
• nprobe: Test case, containing an eBPF probe for measuring the impact of extracting the exact information extracted from nProbe within eBPF, in terms of number of processed packets and memory used
• swap: Test case, containing an eBPF probe for measuring the impact of requesting a snapshot-access to the eBPF maps, both in terms of compilation time and performance degradation while swapping in-out the underlying eBPF programs.
• moongen.lua: used by the client to generate traffic at line rate using MoonGen
• plotter.ipynb: a Python notebook for plotting the results of the tests leveraging matplotlib
• set_irq_affinity.sh: a Bash script for setting the number of CPU used for handling incoming network traffic to 1 for a specific interface

Assumptions

This figure represents the set-up used for the tests, and the following assumptions have been made:

1. Server is the machine that receives network traffic and runs eBPF probes
2. Server supports XDP_DRV mode. Change it in the __main__.py files of the desired test
3. Client-Server have their ssh keys registered under the home folder of the user used for logging-in.
4. Client has a working MoonGen build under the following directory ~/Moongen/build. Change it if needed.

Usage

To run one of the tests between the available ones (adaptiveness, erase, nprobe, swap):

1. go to the root of the project (the current directory)
2. python -m <test_name> --help to read the available arguments
3. python -m <test_name> ... to run the test with the provided arguments

A result file will be available once finished.

Acknowledgements

If you are using OpportunisticMonitoring's code for a scientific research, whether to replicate experiments or try new ones, please cite the related paper in your manuscript as follows:

S. Magnani, F. Risso and D. Siracusa, "A Control Plane Enabling Automated and Fully Adaptive Network Traffic Monitoring With eBPF," in IEEE Access, vol. 10, pp. 90778-90791, 2022, doi: 10.1109/ACCESS.2022.3202644.

I sincerely thank my Ph.D. advisor Domenico Siracusa, and my M.Sc. thesis supervisor prof. Fulvio Risso.