Datasets on Satellite Computing with COTS Devices

In this repository, we release the dataset and tools in MobiCom '24 paper, Deciphering the Enigma of Satellite Computing with COTS Devices: Measurement and Analysis. Amidst the rapid growth of low-Earth orbit constellations, utilizing COTS computing devices as in-orbit computing units offers significant advantages: low cost, high performance, and reusability of software and hardware ecosystems. This paper aims to explore how to better utilize computational resources in orbit, considering the constraints of temperature and power consumption on satellite platforms.

The dataset from this paper is derived from the BUPT-1 satellite, the flagship 12U CubeSat of the TianSuan Constellation project. BUPT-1 equipped with popular COTS computing payloads including Atlas 200 DK and Raspberry Pi 4B, primarily for satellite computing research.

Dataset Structure

The data can be categorized into platform data (sensor-acquired telemetry data within the satellite) and payload data (generated from COTS computing payloads during computational tasks).

The repository is structured around experimental types mentioned in the paper, with folders dedicated to temperature control and energy management. There are five folders for temperature control: Temperature-Overview, Temperature-Power-Variations, Temperature-Overheating, Temperature-HeatingRate, and Temperature-DaylightEclipse. For energy management, the folders are: Energy-Overview, Energy-ShortTerm, Energy-LongTerm, Energy-Available, and Energy-Efficiency. Each folder contains a README providing detailed explanations.

Source code for data processing and plotting, along with the datasets used in the experiments, are contained within each folder. We included the dataset file under the Data folder for all experiments. All telemetry data collected during the experiment is compiled into a table and placed in the CommonData-Telemetries folder for convenience.

Installation Instructions

Usage Instructions

Before you start, you may use these commands to install the Python environment required:

# Makesure $PWD is the root directory of the repo
# Replace "your_env_name" at your ease
conda create -n your_env_name python=3.9.11
conda activate your_env_name
pip3 install numpy pandas scipy matplotlib

Additionally, you should uncompress the dataset file：

• You must use cd navigate to sub-directories in order to run the code.
• Before running code, you must navigate to CommonData-Telemetries folder to unzip the "telemetry_all.csv.zip".

After that, you can run all the scripts in once using run_all.sh:

bash run_all.sh

It would produce all the .pdf and .log results, you may use this command to inspect:

find . -type f \( -name "*.pdf" -o -name "*.log" \)

And the correct results are like:

./Temperature-HeatingRate/sat_pi.log
./Temperature-HeatingRate/gnd_atlas.log
./Temperature-HeatingRate/sat_atlas.log
./Temperature-HeatingRate/gnd_pi.log
./Energy-Available/figure14.pdf
./Temperature-Overview/figure2.pdf
./Temperature-DaylightEclipse/figure8.pdf
./Temperature-DaylightEclipse/figure9.pdf
./Temperature-DaylightEclipse/figure7.pdf
./Energy-ShortTerm/figure11.pdf
./Energy-Overview/figure10.pdf
./Energy-LongTerm/figure13.pdf
./Energy-LongTerm/figure12.pdf
./Energy-Efficiency/sat_pi.log
./Energy-Efficiency/gnd_sat_atlas_cl_dp.log
./Energy-Efficiency/gnd_atlas.log
./Energy-Efficiency/sat_atlas.log
./Energy-Efficiency/gnd_pi.log
./Temperature-Overheating/figure6.pdf
./Temperature-Overheating/figure4.pdf
./Temperature-Overheating/figure5.pdf
./Temperature-Power-Variations/figure3.pdf

We also provide a cleaning scripts to delete all the generated results. Simply run the commands like:

bash clear all.sh

---

Paper Structure to Folder Structure

Content in Paper	Folder in Repo.	Description
Figure 2 (Section 3)	Temperature-Overview	Surface Temperature Varitions with or without Computing Tasks.
Figure 3 (Section 3.1)	Temperature-Power-Variations	Temperature and Power Varitions under different CPU/NPU/Power Level Settings.
Figure 4 to 6 (Section 3.2)	Temperature-Overheating	Overheating Effects on Atlas and Pi.
Table 3 (Section 3.3)	Temperature-HeatingRate	Relations of Power and Heating Rate.
Figures 7 to 9 (Section 3.4)	Temperature-DaylightEclipse	Impact of Daylight and Eclipse Zones on Temperature
Figures 10 (Section 4)	Energy-Overview	Solar/Consumed Power Varitions with or without Computing Tasks.
Figures 11 (Section 4.1, 4.2)	Energy-ShortTerm	Typical Solar/Consumed Power Varitions within an Orbiting Period (about 90 mins).
Figures 12 and 13 (Section 4.1, 4.2)	Energy-LongTerm	Solar/Consumed Energy Varitions within 2 weeks.
Figures 14 (Section 4.2)	Energy-Available	Solar/Consumed Energy Varitions on Identical Time Range.
Table 4 and 5 (Section 4.3)	Energy-Efficiency	Explore the underutilization of solar energy and the excessive use of battery power.

---

Measurement tools

In the terrestrial tests, we use Monsoon Power Monitor to record the current.

Support

If there are any questions, feel free to reach out to us (xrl@bupt.edu.cn)!