Cold Start Latency Mitigation Mechanisms in Serverless Computing: Taxonomy, Review, and Future Directions

Today, Function-as-a-Service (FaaS), as an emerging path to implementing serverless computing paradigm, seems quite promising in responding to the growing increase in requests for microservices. However, the serverless computing model also brings new challenges. Although the execution time of the functions is very short in FaaS, the initialization of containers and cold start is time-consuming and leads to a decrease in system performance. Cold start latency reduces the attractiveness of FaaS with the cost it imposes on the system. This is the most serious challenge in the majority of serverless computing platforms. Despite the importance of the subject, to the best of our knowledge, there is no comprehensive systematic review of cold start latency mitigation mechanisms in serverless computing platforms. This study aims to provide a systematic literature review on cold start latency mitigation mechanisms in serverless computing environment. In our proposed taxonomy, the mechanisms are classified into four categories: application-based, checkpoint-based, invocation time prediction-based, and cache-based; then, every category is introduced and reviewed. Finally, we propose open issues and conclude.