Agile Latency Estimation for a Real-time Service-oriented Software Architecture

Abstract This paper presents our testbed and software pipeline for automatic latency estimation for a service-oriented software architecture (SOA). This type of architecture consists of modular services that are dynamically combined at runtime to form a functioning system. As different service combinations become possible at runtime, agile approaches for testing the resulting systems become necessary. Besides other factors, latencies are of particular interest for the implementation of control systems. Our agile approach automatically generates dummy services, including interfaces and tasks for internal processing, based on a service description in a human-readable format. Services are then automatically distributed to the computers of our testbed, which are connected through Ethernet. We empirically obtain latency estimates for processing and communication steps for a given composition of services. In this paper we describe our data format, abstractions about internal run-time behavior of services and the code generation pipeline. The evaluation presents latency estimates that we are able to obtain through our testbed that resembles the sense-plan-act paradigm.