Making Sense of Constellations: Methodologies for Understanding Starlink's Scheduling Algorithms

Hammas Bin Tanveer,Mike Puchol,Rachee Singh,Antonio Bianchi,Rishab Nithyanand

2023-07-02

Abstract:Starlink constellations are currently the largest LEO WAN and have seen considerable interest from the research community. In this paper, we use high-frequency and high-fidelity measurements to uncover evidence of hierarchical traffic controllers in Starlink -- a global controller which allocates satellites to terminals and an on-satellite controller that schedules transmission of user flows. We then devise a novel approach for identifying how satellites are allocated to user terminals. Using data gathered with this approach, we measure the characteristics of the global controller and identify the factors that influence the allocation of satellites to terminals. Finally, we use this data to build a model which approximates Starlink's global scheduler. Our model is able to predict the characteristics of the satellite allocated to a terminal at a specific location and time with reasonably high accuracy and at a rate significantly higher than baseline.

Networking and Internet Architecture

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

This paper aims to reveal the scheduling algorithms and their working principles within the Starlink satellite network. Specifically, the researchers discovered a hierarchical traffic controller in the Starlink network through high-frequency and high-precision measurement data and conducted an in-depth analysis of it. The main objectives include:

1. **Identify the Global Controller**: The researchers discovered a global controller responsible for assigning satellites to user terminals. This controller performs a global allocation every 15 seconds.
2. **Identify the Satellite Controller**: The researchers also discovered a local controller located on the satellite, responsible for scheduling data streams transmitted from user terminals.
3. **Understand the Allocation Mechanism**: Through data analysis, the researchers revealed the factors influencing the allocation of satellites to user terminals, including geographical location, satellite load, sunlight exposure, etc.
4. **Establish a Model**: The researchers developed an approximate model that can predict the characteristics of satellites allocated to user terminals at specific locations and times.

### Main Contributions

1. **Discovery of Hierarchical Traffic Control**: Through high-frequency measurements, the researchers revealed for the first time the existence of a hierarchical traffic control mechanism in the Starlink network.
2. **Development of New Technology**: The researchers developed a new technology that can identify the satellites allocated to user terminals.
3. **Revealing the Characteristics of the Global Scheduler**: The researchers revealed the characteristics and preferences of the Starlink global scheduler, such as selecting newer satellites, those exposed to sunlight, located in the northwest direction, and at higher angles.
4. **Establishing an Approximate Model**: The researchers developed an approximate model that can predict the characteristics of satellites allocated to user terminals with high accuracy.

### Research Background

Starlink is a low Earth orbit (LEO) satellite constellation consisting of nearly 4000 satellites. Its ecosystem includes in-orbit satellites, user terminals, ground stations, and points of presence (PoP). User terminals provide internet connectivity by connecting to satellites, which then transmit data to the internet backbone through ground stations.

### Experimental Setup

The researchers used four Starlink terminals located in different regions for measurements. These terminals were located in Western Europe, the Northeastern United States, the Midwestern United States, and the Northwestern United States. To improve measurement accuracy, the researchers configured the Starlink routers in bridge mode and connected them to dedicated Raspberry Pis via Ethernet. The destination servers were located at the Starlink PoPs corresponding to the user terminals to minimize the impact of the ground network on the measurement results.

### Observations

1. **Global Controller**: The researchers found that Starlink relies on a global controller to schedule connections between satellites and user terminals. Every 15 seconds, the global controller reallocates satellites.
2. **Satellite Controller**: The researchers also found that the local controller on the satellite is responsible for scheduling data streams from user terminals. The allocation of these data streams is similar to round-robin scheduling.

### Methods

1. **Obtaining Satellite Allocation**: The researchers used Starlink's obstruction map and compared public satellite position data with the obstruction map recorded by the terminals to identify the satellites allocated to specific user terminals.
2. **Feature Selection**: The researchers selected multiple features, including the azimuth, elevation, age, and sunlight exposure of the satellites, to construct the scheduler model.
3. **Training and Validating the Model**: The researchers used a random forest model for training and validated the robustness of the model through five-fold cross-validation and holdout set validation.

### Conclusion

Through high-frequency and high-precision measurements, this paper revealed for the first time the hierarchical traffic control mechanism in the Starlink network and established an approximate model to predict satellite allocation. These findings contribute to future modeling and evaluation of the Starlink network, providing a foundation for further optimization of network performance.