Keynote: Designing Serverless Platforms to Support Emerging Applications

Serverless computing offerings from cloud providers have gained significant traction in recent years due to the advantages that these platforms bring with their flexible pricing models, built-in scalability, and minimal operational requirements. In a recent survey of serverless use cases, we found a wide variety of applications that depend on these services, including implementing the core functionality at the backend of mobile applications, automating the DevOps tasks of complex distributed applications, real-time processing of IoT streaming data, and scientific applications. To properly support these applications, the platforms should be fast, self-managing, and provide support for diverse QoS requirements. As a result, novel improvements to serverless platforms are rapidly being proposed and adopted. Evaluating these solutions necessitates application-based, workload-aware benchmarking tools that the community can rely on. This talk addresses these challenges and our research efforts on tackling them, presenting a performance engineering perspective about the current state and future challenges of serverless computing research. I will describe our solutions in autonomic resource management for serverless platforms, focusing on solutions that improve performance or reduce costs via scheduling, caching, and right-sizing of resources, along with our ongoing efforts in developing an application-driven serverless benchmark.