Welcome to the official GitHub organization for Stimpack, a novel system presented at NSDI '26.

Stimpack is a research-driven framework designed to solve the scalability bottleneck of cloud gaming at the edge. By identifying that high-quality rendering is often wasted over lossy networks, Stimpack adaptively balances server-side rendering costs against user-perceived quality to maximize system-wide utility.

The Stimpack framework is divided into two main pillars. To fully experience or reproduce our research, you will need both the "Brain" (Runtime) and the "Body" (Plugin).

The Orchestration Engine.

• Role: Handles multi-user RQ optimization, efficiency scoring, and VMAF-based quality prediction.
• Tech Stack: Python 3.10+, uv, ZeroMQ.
• Core Feature: Contains pre-trained ML models that estimate user-side visual quality without expensive pixel-level comparisons.

The Execution & Environment.

• Role: Integrates into Unreal Engine 5 to monitor real-time rendering telemetry and apply RQ adjustments.
• Tech Stack: Unreal Engine 5.0.3, C++, ZeroMQ.
• Core Feature: Includes three production-ready game environments used in the official NSDI evaluation.

• 2x Scalability: Serves twice as many concurrent users on the same GPU resources compared to static baselines.
• 24% Quality Improvement: Maximizes overall service quality by prioritizing resources where they matter most.
• Resource-Aware: Seamlessly balances FPS (SLO) and visual fidelity.

The project name is a tribute to the Terran Marine's iconic ability in Blizzard's StarCraft series.

As a former professional gamer, the lead author designed the system with the same high-stakes efficiency in mind: trading off "invisible" rendering details (health) for a massive boost in system-wide combat speed and scalability.

If you use Stimpack in your research, please cite our paper:

@inproceedings{heo2026stimpack,
  title={Stimpack: An Adaptive Rendering Optimization System for Scalable Cloud Gaming},
  author={Heo, Jin and Wang, Vic and Bhardwaj, Ketan and Gavrilovska, Ada},
  booktitle={23rd USENIX Symposium on Networked Systems Design and Implementation (NSDI 26)},
  year={2026}
}

Visit author's webpage | Contact the Authors