Fast-forward runtime monitoring - an industrial case study

Abstract

Amongst the challenges of statefully monitoring large-scale industrial systems is the ability to efficiently advance the monitors to

the current state of the system. This problem presents itself under various guises such as when a monitoring system is being deployed for the

first time, when monitors are changed and redeployed, and when asynchronous monitors fall too much behind the system.

We propose fast-forward monitoring — a means of reaching the monitoring state at a particular point in time in an efficient manner, without

actually traversing all the transitions leading to that state, and which we applied to a financial transaction system with millions of transactions already affected. In this paper we discuss our experience and present a generic theory of monitor fast-forwarding instantiating it for efficient monitor deployment in real-life systems.