Poster: Local Self-Adaptation for Distributed Complex Event Processing

Complex Event Processing (CEP) systems continuously evaluate query workloads over event streams to detect patterns of interest. In networks of event sources, in-network processing methods optimize processing metrics by constructing evaluation plans based on network statistics, which dictate the placement of operators and forwarding of results. However, changing network statistics can reduce the effectiveness of an evaluation plan, necessitating its adaptation. Existing approaches for adaptive distributed CEP often rely on redeploying the entire evaluation plan or support only limited local adaptations, typically requiring central coordination and network-wide information. We argue that these measures are inefficient when changes affect only parts of the evaluation plan or are transient. To address this, we advocate for local self-adaptation. We illustrate a local self-adaptation strategy that operates in a fully decentralized manner, effectively mitigating the degradation of an evaluation plan optimized for reducing network communication when network rates change. Building on this, we identify requirements for a framework that enables the definition of local self-adaptation strategies for distributed CEP.