A model-driven approach for the formal verification of storm-based streaming applications

Abstract

Data-intensive applications (DIAs) based on so-called Big Data technologies are nowadays a common solution adopted by IT companies to face their growing computational needs. The need for highly reliable applications able to handle huge amounts of data and the availability of infrastructures for distributed computing rapidly led industries to develop frameworks for streaming and big-data processing, like Apache Storm and Spark. The definition of methodologies and principles for good software design is, therefore, fundamental to support the development of DIAs. This paper presents an approach for non-functional analysis of DIAs through DVerT, a tool for the architectural assessment of Storm applications. The verification is based on a translation of Storm topologies into the CLTLoc metric temporal logic. It allows the designer of a Storm application to check for the existence of components that cannot process their workload in a timely manner, typically due to an incorrect design of the topology.