Data stream processing (DSP) has emerged over the years as the reference paradigm for the analysis of continuous and fast information flows, which often have to be processed with low-latency requirements to extract insights and knowledge from raw data. Dealing with unbounded dataflows, DSP applications are typically long running and thus, likely experience varying workloads and working conditions over time. To keep a consistent service level in face of such variability, a lot of effort has been spent studying strategies for runtime adaptation of DSP systems and applications. In this survey, we review the most relevant approaches from the literature, presenting a taxonomy to characterize the state of the art along several key dimensions. Our analysis allows us to identify current research trends as well as open challenges that will motivate further investigations in this field.
1 October 2018 in Future Generation Computer Systems
Data Stream Processing (DSP) applications analyze data flows in near real-time by means of operators, which process and transform incoming data. Operators handle high data rates running parallel replicas across multiple processors and hosts. To guarantee consistent performance without wasting resources in the face of variable workloads, auto-scaling techniques have been studied to adapt operator parallelism at run-time. However, most of the effort has been spent under the assumption of homogeneous computing infrastructures, neglecting the complexity of modern environments.We consider the problem of deciding both how many operator replicas should be executed and which types of computing nodes should be acquired. We devise heterogeneity-aware policies by means of a two-layered hierarchy of controllers. While application-level components steer the adaptation process for whole applications, aiming to guarantee user-specified requirements, lower-layer components control auto-scaling of single operators. We tackle the fundamental challenge of performance and workload uncertainty, exploiting Bayesian optimization (BO) and reinforcement learning (RL) to devise policies. The evaluation shows that our approach is able to meet users’ requirements in terms of response time and adaptation overhead, while minimizing the cost due to resource usage, outperforming state-of-the-art baselines. We also demonstrate how partial model information is exploited to reduce training time for learning-based controllers.