Complex event recognition in the Big Data era: a survey

Abstract

The concept of event processing is established as a generic computational paradigm in various application fields. Events report on state changes of a system and its environment. Complex Event Recognition (CER) refers to the identification of composite events of interest, which are collections of simple, derived events that satisfy some pattern, thereby providing the opportunity for reactive and proactive measures. Examples include the recognition of anomalies in maritime surveillance, electronic fraud, cardiac arrhythmias and epidemic spread. This survey elaborates on the whole pipeline from the time CER queries are expressed in the most prominent languages, to algorithmic toolkits for scaling-out CER to clustered and geo-distributed architectural settings. We also highlight future research directions.

A Compositional Approach for Complex Event Pattern Modeling and Transformation to Colored Petri Nets with Black Sequencing Transitions

“…For example, the color set INT denotes the set of integer numbers, REAL denotes the set of real numbers, STRING denotes the set of strings, etc. Furthermore, new color sets can be defined as a Cartesian product, a record 2 or the union of existing color sets. For instance, color set INT2 = product INT × INT is defined as the Cartesian product of two integers, whereas RI2 = record f1 : INT * f2 : STRING is a record with two fields f 1 and f 2, and the 2.…”

“…Complex Event Processing (CEP) technology [1,2] is used in many different fields to analyze large volumes of information in the form of streams of events, using them to take decisions or derive conclusions. Areas of application include industry and business [3], sensors and IoT [4], air quality control [5], etc.…”

Complex Event Processing (CEP)

“…They can limit the evaluation time or the number of tuples which are processed by an otherwise blocking or stateful operator. Other methods are, for example, compensation-based techniques, where operators in the query are executed in the same way as if all events would be ordered or approximation-based techniques, where either streams are summarized and events are approximated or approximation is done on a recent history of events (Giatrakos et al 2019).…”

“…Besides load balancing and load shedding, parallelization techniques can be applied to increase the performance of a system. Giatrakos et al (2019) distinguish two kinds of parallelization for CEP, namely, task and data parallelization, where task parallelization comprises the distribution of queries and subqueries or single operators to different nodes where they are executed. Data parallelization considers the distribution of data to multiple instances of the same operator or query.…”

“…A possible strategy to implement CEPs in a scalable way is elevating CEP systems to big data platforms as these are designed to serve high workloads by workload distribution and elastic services. Giatrakos et al (2019) show how CEP can be integrated with Spark Streaming, Apache Flink, and Apache Storm taking advantage of the corresponding abilities for scalability.…”