1. Tetris is a serverless platform designed to reduce memory consumption for deep learning inference services.
2. Tetris minimizes runtime redundancy through batching and concurrent execution, and eliminates tensor redundancy across instances from different functions using a lightweight and safe tensor mapping mechanism.
3. Evaluation shows that Tetris can save up to 93% memory footprint for inference services, and increases the function density by 30× without impairing the latency.
The article is generally reliable in terms of its content, as it provides detailed information about the design of Tetris, its features, and its evaluation results. The authors provide evidence to support their claims, such as the evaluation results showing that Tetris can save up to 93% memory footprint for inference services. However, there are some potential biases in the article that should be noted. For example, the authors do not explore any counterarguments or alternative solutions to the problem they are addressing with Tetris. Additionally, they do not discuss any potential risks associated with using Tetris or other serverless platforms for deep learning inference services. Finally, while USENIX is committed to Open Access to research presented at their events, it is unclear if this article has been peer-reviewed or not.