CAST: a Declarative Language and its Execution Platform for Large-Scale Cloud Simulations

Cloud computing has become increasingly popular amongst organizations as a cost-effective solution to serve large-scale applications. Cloud deployments, however, are undermined by a fundamental disconnect between the software development process and the cloud environment: testing of applications on the cloud is expensive, and poorly optimized software often results in poor performances and functional issues when scaled up in the cloud. In this scenario, cloud simulators are promising tools that let engineers and researchers test applications and novel algorithms using limited resources. Current state of the art simulators, such as CloudSim, leverage outdated software frameworks that force developers to go through a steep learning-curve and did not keep up with the scale of today’s real-world deployments which involve thousands of servers distributed across multiple regions. This work addresses the shortcomings of current cloud simulation tools and introduces a novel, distributed, scalable, and extensible cloud simulation framework: CAST. CAST brings cloud simulations to the cloud by leveraging the actor model to distribute computation efficiently. Moreover we designed CASTDSL, a declarative language that allows engineers to write simulations by providing their high-level description.