Intelligent Boosting for Low Power-High Performance System

In today's era, where we are perpetually surrounded by technology, smartphones have become the center of our digital lives. Even with the constant influx of new features, users expect the best mobile performance. One of the most popular ways that original equipment manufacturers (OEMs) use to improve app launch is CPU Boosting. Existing solutions that use this method increase CPU core frequency for a fixed amount of time, leading to either excessive power consumption or subpar launch performance. This paper introduces Balanced Booster, an ML based platform that uses knowledge of the current system condition and similar apps' behavior to predict the optimal duration for boosting, thus ensuring the best performance and minimal power usage. For resource intensive apps like camera and games, this prediction also includes the duration after launch for which the app would need boosting to support its heavy operations. With Balanced Booster, we predict optimal boosting time for launch with 92% accuracy, demonstrating a 6% reduction in app launch time with battery lasting for up to 2 hours more compared to base scenario.