- Competition Page
Architecture
GitWatt: A Journey into Cloud Optimization and Sustainability
In this year(2024.07~08), I participated in the Google Gemini API Developer Competition with a product concept called GitWatt with my co-workers. The goal was simple yet ambitious: to recommend the most optimal cloud resources for running your code while also providing insights into electricity consumption and environmental impact. Here's a review of the journey and the vision behind GitWatt.
The Idea Behind GitWatt
As an engineer, I've often wondered:
- Am I using the most cost-efficient and energy-efficient cloud resources for my workloads?
- How much electricity does my code consume, and what is its environmental impact?
GitWatt was born out of the desire to answer these questions. By leveraging the power of LLMs (Large Language Models) and APIs like Google Gemini, the idea was to create a tool that:
- Analyzes your Git repository to understand the code structure and resource requirements.
- Recommends cloud resources (e.g., AWS, GCP, Azure) that balance performance, cost, and environmental sustainability.
- Calculates power consumption and estimates carbon emissions for running your code.
The ultimate vision was to enable developers not only to optimize their code but also to make environmentally conscious decisions—all integrated seamlessly with GitHub.
The Core Features
- Cloud Resource Optimization:
- By analyzing the structure of your Git repository and the types of tasks your code performs, GitWatt could recommend the most suitable cloud resources. For example, it could suggest the most cost-efficient instance types across different cloud providers.
- Electricity Consumption Tracking:
- GitWatt calculated how much power your code would consume when run on a specific instance type.
- Using IP-based location analysis, it even factored in regional electricity costs and carbon intensities to provide a detailed breakdown.
- Environmental Impact Assessment:
- By estimating the carbon footprint of running your code, GitWatt highlighted the trade-offs between performance and sustainability.
- Real-Time Feedback:
- A web-based UI provided a futuristic, AI-powered experience for developers. Input your GitHub repository and branch, and see real-time results on cost, power usage, and environmental impact.
Scoring and Evaluation
GitWatt adopted a scoring system that emphasized:
- Cost Efficiency (40 points): Monthly cloud costs were normalized across providers to ensure fair comparisons.
- Performance (30 points): Evaluated based on CPU, memory, and GPU utilization.
- Environmental Impact (30 points): Lower carbon footprints scored higher.
This system allowed developers to quickly see how their code performed across different dimensions and make informed decisions.
Future Vision for GitWatt
While GitWatt focused on cost and environmental optimization during the competition, the potential applications extend far beyond:
- Code-Level Optimization: By integrating directly with GitHub or IDE extensions, developers could receive real-time suggestions for improving the efficiency of their code.
- Private Code Analysis: Exploring how private LLMs could securely analyze private Git repositories for similar insights.
- Environmental Awareness: Encouraging a shift toward green computing by providing actionable insights into the carbon footprint of software development.
Ultimately, I hope GitWatt (or tools like it) can pave the way for a more sustainable and efficient software development ecosystem.
Reflections on the Competition
Participating in the Google Gemini API Developer Competition was an enriching experience. It reinforced the importance of balancing performance with sustainability and the need for tools that empower developers to make better decisions. While GitWatt is still in its early stages, the feedback and interest it received have been incredibly motivating.
If you're interested in sustainable software development, I'd love to hear your thoughts. Let's build a future where our code doesn't just perform better—it leaves a lighter footprint on the planet.