A First Look at WebAssembly Performance in Satellite Computing

This repository contains the evaluation tool SatelliteEval and a benchmark suite developed to study WebAssembly (Wasm) performance in satellite computing. It supports multiple runtime environments, execution modes, and representative satellite onboard tasks, enabling reproducible performance measurements.

SatelliteEval is designed to systematically collect performance data such as execution latency, energy consumption, and thermal impact under simulated satellite conditions. The benchmark suite includes tasks spanning scientific computation, data processing, and AI inference.

Repository Contents

• SatelliteEval/ — Evaluation tool for orchestrating experiments on various Wasm and container-based runtimes.

  • Components: Planning, Deployment, Execution, Logging, Extraction

  • Supported Runtimes:

    • WasmEdge (AOT, JIT, Interpretation)
    • Wasmtime (AOT, JIT)
    • WAMR (AOT, JIT, Interpretation)
    • Wasmer (AOT, JIT)
    • containerd (baseline container runtime)

  • Execution Modes: Ahead-of-Time (AOT), Just-In-Time (JIT), Interpretation

• benchmarks/ — Suite of representative satellite onboard tasks:

  Category	TaskID	Source Language	Description
  Simple Example	Task1	C++, Rust, Go	Prints a greeting message
  Scientific Computing	Task2	C++, Rust, Go	Generates Fibonacci numbers
  Scientific Computing	Task3	C++, Rust, Go	Approximates definite integrals
  Scientific Computing	Task4	C++, Rust, Go	MergeSort on arrays
  Scientific Computing	Task5	C++, Rust, Go	Builds and checks binary trees
  Data Processing	Task6	C++, Rust, Go	Median filtering on images
  Data Processing	Task7	C++, Rust, Go	LZW compression/decompression
  Data Processing	Task8	C++, Rust, Go	Fast Fourier Transform
  AI Inference	Task9	Rust	Image classification (MobileNetV2)
  AI Inference	Task10	Rust	Object detection (YOLOv4)

Getting Started

Prerequisites

• Raspberry Pi 4B (or comparable COTS device) or equivalent x86/ARM host for testing.
• Supported runtimes installed (see list above).

Running SatelliteEval

Planning Component

The Planning component (planning_component/planner.py) generates experiment command lines based on a YAML configuration file (config.yaml).
It reads task IDs (corresponding to container_task_list.tsv or wasm_task_list.tsv), runtime (wasm or container), and parameter mappings of tasks to produce a complete command string for the Deployment component.

Usage:

cd SatelliteEval/planning_component
python3 planner.py --config config.yaml --output command.txt

Deployment Component

The Deployment component is responsible for transmitting and executing experiment commands generated by the planning component.
It provides two deployment modes:

• command_coding.sh — Encodes and decodes commands into Raspberry Pi control frames for satellite communication scenarios.
  Used when commands are transmitted across actual satellite or remote links.

• deploy_local.sh — Executes commands locally without transmission.
  Used for single-device or simulation environments.

Usage (local execution):

cd SatelliteEval/deployment_component
bash deploy_local.sh --input ../../planning_component/command.txt --verbose

Usage (satellite transmission):

# To encode:
bash command_coding.sh --input ../../planning_component/command.txt --output frame_45pi.hex
# To decode:
bash command_coding.sh --output frame_45pi.hex --reverse

Execution Component

The Execution component orchestrates benchmark execution across Wasm and container environments.
It loads paths and parameters from config.yaml, dispatches tasks via test.sh, and records performance data and logs.
All runtime-specific operations (e.g., Wasm or container) are handled by wasm_test.sh and container_test.sh, , while temperature monitoring is managed by temp_logger.sh.

Usage:

cd SatelliteEval/execution_component
bash test.sh <runtime> <interval> <repeat> <clear_report:y|n> <with_input:y|n> <task_id1> [param1] ...

Example:

bash test.sh wasm 2 3 y n 1 2 3

Logging Component

The Logging component collects runtime telemetry data during benchmark execution.

• temp_logger.sh — Continuously records device temperature every 1 s from the onboard thermal sensor.
  It is automatically launched by the Execution component and does not require manual invocation.

• start_HPVM.py — Records power data (voltage and current) via a connected Monsoon High-Voltage Power Monitor.
  It runs on the host computer connected to the power monitor through USB and logs samples to power_log.csv.

Extraction Component

The Extraction component (extractor.py) processes and aggregates experiment results after execution.
It merges performance reports, temperature logs, and power monitor data to compute metrics such as average execution time, energy consumption, and thermal characteristics for each task.

Usage:

cd SatelliteEval/extraction_component
python3 extractor.py config.yaml