Cloud Function Performance: a component modeling approach

Cloud Functions are a trend in cloud computing in which developers are allowed to install code in a Function-as-a-Service (FaaS) platform able to manage provisioning, execution, monitoring and automatic scaling. The underlying infrastructure in FaaS platforms is hidden from the developers and designers and, since the inuence of the infrastructure is unknown, this makes it di_cult to apply software performance engineering approaches on cloud functions, which could lead to wrong or inaccurate performance estimations. In this study, we explore the use of component-based modeling and simulation in order to generate performance estimations of an exemplar cloud function which was exercised using a variety of workloads. A cloud function was both implemented and instrumented to record performance datain a log _le, associated with its invocations; using the log _le as an input, we extracted a performance model in a Palladio Component Model format suitable for running simulations to validate whether the generated model could explain the runtime behavior of the function. Using this approach and further tunings in the model, we were able to validate that the simulations could explain more than 95% of the function's behavior and that component-based modeling and simulation can be considered a serious option when trying to explain the behavior of a cloud function.

Towards autoscaling of Apache Flink jobs

Data stream processing has been gaining attention in the past decade. Apache Flink is an open-source distributed stream processing engine that is able to process a large amount of data in real time with low latency. Computations are distributed among a cluster of nodes. Currently, provisioning the appropriate amount of cloud resources must be done manually ahead of time. A dynamically varying workload may exceed the capacity of the cluster, or leave resources underutilized. In our paper, we describe an architecture that enables the automatic scaling of Flink jobs on Kubernetes based on custom metrics, and describe a simple scaling policy. We also measure the e ects of state size and target parallelism on the duration of the scaling operation, which must be considered when designing an autoscaling policy, so that the Flink job respects a Service Level Agreement.