The complete thesis paper can be downloaded here.
The runtime verification tool MonPoly with its extensive temporal specification language MFODL provides a broad set of features, including real-time monitoring, global quantification, regular expressions matching over program traces, and computing SQL-like aggregations. While JSON-formatted application logs have become ubiquitous, allowing systems to output events as arbitrary complex data structures, both MonPoly and MFODL lack support for custom event data types.
To rectify this, we present an extension of MFODL that allows users to formulate policies over events of custom product types. A type-checking algorithm accompanies the extension with full type-inference support. Moreover, we provide an extension of MonPoly that supports monitoring JSON-formatted application logs with no or minimal preprocessing. We accomplish this with a formula compiler and a generalized log stream preprocessor.
Our case study on two real-world scenarios shows that our extension not only simplifies the monitoring process of complex-typed event streams but it allows users to formulate specifications over complex data structures more naturally, increasing the readability and comprehensibility of monitoring policies.
The source code of this thesis is available on Github: https://github.com/remolueoend/monpoly.