The Case for a New Cloud-Native Programming Model with Pure Functions

Cloud evolution: Over the past two decades, the cloud has become the dominant platform for running all kinds of applications, from data analytics to web services. In the process, cloud platforms have evolved from renting virtual machines (VMs) on-demand to offering elastic compute and storage services. While the ability to support legacy applications was critical in the early days of cloud to ease migration from on-premise, today's users commonly develop cloud-native applications by composing cloud storage services (e.g., S3), compute services (e.g., AWS Lambda), data analytics services (e.g., BigQuery), machine learning services (Azure ML), and elastic databases (e.g., Snowflake [4]). With this approach, users no longer need to explicitly provision CPU/memory/storage for their applications, as the elastic services automatically scale-out based on load and bill users for the resources consumed [7].