A Digital Twin (DT) is a comprehensive digital representation of an actual system, service or product (the Physical Twin, PT), synchronized at a specified frequency and ifdelity [ 1]. The digital twin includes the properties, condition and behavior of the physical entity through models and data, and is continuously updated with real-time data about the PT performance, maintenance, and health status throughout its entire lifetime [2]. The exchange of data between the digital and the physical twins takes place through bi-directional data connections. Additionally, a DT system can also comprise a set of services that permit exploiting the data exchanged by the two twins [3].
1. Introduction France
Alternative solutions use time series databases, such as InfluxDB or TimescaleDB, and
even define temporal models for handing them at a higher level of abstraction [
In this work we explore the use of Graph databases [
We have implemented a data lake using Neo4j, and evaluated its performance and expressiveness against our previous implementation with Redis. As expected, Neo4j allows easy specification of more complex queries using its Cypher query language. The same queries in Redis require the addition of ad-hoc records with the corresponding key-value mappings if the queries were not contemplated beforehand in the Redis record structure. Interestingly, the additional queries forced by these new records introduce a performance penalty that makes Neo4j’s response times better than those of Redis. Furthermore, the response times obtained for simple queries in Redis are not very diferent from those of Neo4j. All this makes Neo4j appear to be a better solution than Redis for realizing data lakes. The description of the tests carried out and their results are availble from https://github.com/atenearesearchgroup/dt-graph-database. As future work, we are defining a benchmark with diferent types of queries that will be used to compare implementations of data lakes using diferent technologies, including time series databases, too. We hope that our evaluation will help to shed some light on the advantages and limitations of each solution, and to identify situations where one type of solution outperforms the other.