Scalability of composite digital twins on OpenTwins platform

Abstract

This paper explores the scalability limitations faced by the OpenTwins digital twin platform, especially when trying to define a digital twin with a high complexity or number of entities. A critical evaluation of the current architecture of the system is performed, identifying problems associated with the base framework, Eclipse Ditto, and the InfluxDB time series database. The use of an actor system and mailboxes by Eclipse Ditto, together with the InfluxDB approach to identify series, is pointed out as a determining factor in the limited scalability of these solutions in terms of cardinality. In order to overcome these constraints, an redefined architecture is proposed that integrates solutions designed to operate with low latency and high cardinality. In this approach, a new component is introduced with the ability to deploy Telegraf instances to efficiently manage digital twin connections. Moreover, as part of the improvement, the time series database is changed, opting for Timescale due to its efficient scalability against high cardinality situations.