Treffer: Title : Cold Start Problem in Serverless Computing: Measurement and Optimization Techniques

Serverless computing, or Function-as-a-Service (FaaS), has become a dominant paradigm in cloud architecture. A significant performance challenge associated with this model is the "cold start” a latency introduced when provisioning new execution environments. This thesis presents a systematic, empirical investigation into cold start latency on the leading serverless platforms: AWS Lambda and Azure Functions. A quantitative methodology was employed for research, which was conducted between October and December 2025. The work began with a literature review of existing mitigation techniques, followed by the design and execution of automated benchmark tests. These tests systematically evaluated the impact of key variables on initialization time, including runtime environment (Python vs. Node.js), memory allocation, and specific optimization strategies such as AWS Provisioned Concurrency and custom warm-up triggers. The primary goal of the research was to quantify cold start latencies under these various configurations and to empirically assess the effectiveness and performance trade-offs of different mitigation strategies. The results are delivered as a comprehensive benchmark dataset, the automated scripts used for data collection, a comparative analysis report, and a set of practical, evidence-based recommendations. This work provides significant value to both the academic community by contributing a modern benchmark dataset, and to industry practitioners by offering clear guidelines for building performance and cost-effective serverless applications.