This is the artifact for Chronos: Scalable and Cost-Effective Cloud-based RAN Emulation System.
Chronos is a cloud-deployable RAN emulation system built around four ideas:
- Bypass the compute- and bandwidth-intensive PHY layer using nFAPI.
- Virtualize time so cloud preemptions and network delays do not cause RAN deadlines to expire inside the emulation.
- Advance components through RAN slots using distributed synchronization barriers.
- Distribute the middlebox that forwards control- and data-plane traffic between gNBs and UEs.
The artifact provides the implementation, Powder/Emulab deployment automation, experiment data, scripts used to generate the paper plots, prepared PDF outputs, and access to a hosted Powder evaluation deployment.
The repository includes
chronos-artifact.mp4, a video tutorial covering the
complete artifact and Powder evaluation workflow.
The artifact supports:
- Artifacts Available: the source, deployment scripts, experiment data, and plotting workflow are publicly available in this repository.
- Artifacts Evaluated—Functional: evaluators can inspect a running Chronos deployment on Powder and verify end-to-end operation.
- Results Reproduced: evaluators can regenerate the paper plots from the supplied data and reproduce the scale-dependent dilation result on Powder.
git clone https://github.com/netsys-edinburgh/EmuRAN-Draft-Artifact.git
cd EmuRAN-Draft-Artifact| Path | Contents |
|---|---|
code/ |
Chronos virtual-time, synchronization, middlebox, gNB, and UE implementations. |
deployment/ |
Powder/Emulab provisioning scripts, Kubernetes Helm charts, and operational scripts. |
results/data/ |
Raw logs, packet captures, processed data, helper modules, and plotting scripts. |
results/output/ |
Prepared and newly generated PDF plots. |
plot.ipynb |
One plotting cell per paper figure or panel. |
requirement.txt |
Python dependencies for the plotting workflow. |
install_dependencies.sh |
Linux/macOS dependency installer. |
ae.tex / ae.pdf |
Full artifact-evaluation instructions. |
- Linux or macOS
- Python 3.10 or newer
- Internet access for the initial dependency installation
- No GPU, LaTeX installation, or privileged access is required
Install all dependencies into a local .venv:
./install_dependencies.shThe installer uses requirement.txt and installs Jupyter, IPython, IPykernel,
NumPy, Pandas, Matplotlib, Seaborn, and PyMuPDF. It also registers the dedicated
Python (Chronos Artifact) notebook kernel. To override the Python executable
or environment location:
PYTHON_BIN=/path/to/python3 VENV_DIR=/path/to/venv ./install_dependencies.shLaunch the notebook from the repository root:
source .venv/bin/activate
jupyter notebook plot.ipynbIf Jupyter prompts for a kernel, select Python (Chronos Artifact). This makes
the notebook use the environment containing PyMuPDF and the other plotting
dependencies.
Each code cell:
- runs the corresponding plotting script under
results/data/; - writes a fresh PDF to
results/output/; and - displays that newly generated PDF inline.
Run one cell for one figure, or select Run All. A complete run produces 30
PDF files in results/output/.
The complete notebook can also be executed non-interactively:
source .venv/bin/activate
jupyter nbconvert --to notebook --execute plot.ipynb \
--output plot.executed.ipynb --ExecutePreprocessor.timeout=300
find results/output -maxdepth 1 -name '*.pdf' | wc -lThe expected PDF count is 30. The notebook uses paths relative to the
repository root, so it can be cloned and run without editing machine-specific
paths.
The authors provide a running Chronos deployment on Powder from September 1 through September 15.
The video tutorial chronos-artifact.mp4 demonstrates
the complete Powder evaluation workflow.
Post your public SSH key in a comment on the paper's HotCRP artifact-evaluation page. The authors will add the key to the deployment and reply through HotCRP with the SSH command, hostname, username, and any required jump-host instructions.
Only send the public key, such as the contents of id_ed25519.pub. Keep the
matching private key, passphrase, passwords, and access tokens private.
After receiving the connection details, connect using the command provided in HotCRP. Its general form is:
ssh -i <private-key> <user>@<host>The hosted Powder experiment reproduces a key paper result: as the number of emulated gNB/UE pairs increases on a fixed compute allocation, Chronos increases its dilation factor. This gives components additional wall-clock time per unit of virtual time and preserves coordinated slot execution when compute becomes constrained.
After connecting to node0 using the SSH command supplied through HotCRP,
connect from node0 to the controller and enter the deployment directory:
ssh ubuntu@10.2.1.2
cd chronos-auto-deployFor each scale in 1, 10, 50, 100, and 200:
- Open
values.yamland set bothnumberGNBandnumberUEto the same scale value. - Run
./run_experiment.shto deploy Chronos. - In the interactive interface, press
pto display the pods. Wait until all expected pods are running. - Run
./core-logs.shand confirm that all gNBs and UEs are connected. - In the interactive interface, enter
l globalscto display the global coordinator log. Observe and record the dilation value. - Run
./teardown-experiment.shand wait for the deployment to be removed. - Change both scale values in
values.yamlto the next value and repeat.
Do not start the next scale until the previous deployment has been torn down.
A dilation factor near 1 means that the emulation is keeping pace with real
time; a value above 1 means Chronos allocates additional wall-clock time to
complete each unit of virtual time.
The result is reproduced when dilation remains near real time while compute capacity is sufficient and then rises as increasing gNB/UE scale creates compute pressure. Record the value observed at every scale; the reproduced result is the increasing dilation trend while coordinated slot execution continues.
| Path | Role |
|---|---|
code/supervisor/ |
Patched Linux 5.15/KVM source that serves guest RDTSC reads from Chronos's virtual TSC. |
code/slot_monitor/ |
custom_tsc module and slot-monitor utilities controlling virtual time. |
code/middlebox/ |
Distributed multi-UE nFAPI proxy and slot-completion logic. |
code/gnb/ |
Modified OpenAirInterface gNB running in nFAPI VNF/emulated-L1 mode. |
code/ue/ |
Modified OpenAirInterface UE running in standalone nFAPI PNF/emulated-L1 mode. |
At runtime, local slot monitors stop virtual time at slot boundaries. The middleboxes and participating components report completion to the global slot coordinator. Once all active components complete the slot, the coordinator releases the next slot and local virtual clocks resume.
deployment/provisioning/profile.py is
the Powder/Emulab profile. It allocates controller/compute, proxy, and global
slot-coordinator nodes and invokes the role-specific provisioning scripts.
deployment/provisioning/scripts/ builds
and installs the patched kernel, configures networking and SCTP, creates nested
Ubuntu/KVM guests, installs virtual-time services, and constructs the
Kubernetes cluster.
deployment/helm/ contains charts for:
| Chart | Component |
|---|---|
core/ |
Mobile core and subscriber database |
globalsc/ |
Global slot coordinator |
proxy/ |
Distributed Chronos middleboxes |
gnb/ |
OAI gNB instances |
ue/ |
OAI UE instances and data-plane measurement sidecars |
persistue/ |
Persistent UE support used by selected experiments |
The central configuration file is
deployment/helm/values.yaml. Operational
scripts in the same directory start UEs, inspect or restart components, change
scale parameters, and collect logs.
See the artifact document for the complete evaluation procedure and
access details. Additional provisioning notes are available in
deployment/provisioning/README.md.