| dc.description.abstract | Cloud computing technology has enabled enterprises in all societal domains to achieve remarkable growth by harnessing data and computation power without an upfront investment in computer infrastructure or any expense in maintaining hardware. In the Infrastructure-as-a-Service model, cloud customers acquire virtual servers, i.e., they gain remote access to the operating system of their choice. Free of any concerns or responsibilities regarding the hardware, customers can focus more on the business and services offered by their applications. However, there is still the need to manage their software stacks, which is often quite complex. Serverless computing is a new cloud paradigm that aims to free software developers from the burden of maintaining their software infrastructure, i.e., the operating system, libraries, and runtime execution environments supporting their applications. The serverless computing paradigm has gained traction and, for many, it represents the future of cloud computing. Academic researchers, industry leaders, and software developers are exploring the opportunities and challenges introduced by serverless computing. This new model presents challenges for all stakeholders. For the service providers leading this space currently (namely, Amazon, Microsoft, Google, and IBM), the new technology requires changes in how computing resources (CPU, memory, network, and storage) are managed in order to achieve the targeted efficiency and economy-of-scale. Academic researchers have been innovating in resource management for decades, having succeeded in improving the efficiency of supercomputers, mobile devices, private enterprise servers, and cloud computing providers. Industry researchers can experiment with new ideas in the context of a company's implementation of serverless computing, having access to the company’s intellectual property. Academic researchers need to rely on open-source implementations of serverless computing infrastructure, often with their progress hindered by deficiencies and limitations in the publicly available alternatives. Even for experienced system software researchers, the initial learning curve is quite steep. In this work, we aim at making the experimentation with serverless computing frameworks practical for researchers with expertise in computer science but lacking experience in distributed systems and container technology. We assume the level of proficiency of a senior undergraduate or an early graduate student. Our research experiments with a popular open-source framework, OpenFaaS, to identify possible pitfalls in its deployment and find the components involved in its resource management strategy. The results of our exploration will simplify the path for introducing research management optimizations. | |
