Advances in Brain-Computer Interfaces and Neural Integration
Seminar Cloud Computing Cold Start Problem in Serverless Computing
Abstract
Dincer Atasoy
This study examines the ’cold start problem’ in serverless computing. Serverless computing is known for its ability to provide services on-demand and cost-effectively. However, it faces a challenge known as cold starts. This occurs when there is a delay in starting serverless functions after they have not been used for some time. This delay can affect the performance and user experience, especially in applications where time is critical. We look at different solutions offered by commercial platforms and academic research. A key focus is on the FaasCache method, which uses a caching technique to keep functions ready and reduce cold starts. We compare this with other methods like pre- warming functions, using different types of servers, and predicting usage patterns with AI. The paper also discusses new challenges, especially when using AI applications with serverless computing, which have their own issues with resource use and start-up times.