As connected devices and services multiply, the way they coordinate becomes an important part of IoT scalability. IoT service interactions determine how control and data move through an application, whether components need to know one another’s locations, and how easily teams can understand and change the workflow. Choosing an interaction pattern is therefore an architectural decision, not simply a choice of messaging technology.
A comparative analysis by Damian Arellanes and Kung-Kiu Lau evaluates four interaction schemas against four scalability considerations: visible control flow, separation of control from computation, decentralization and location transparency. A real-time electrocardiogram monitoring scenario illustrates how the patterns differ. The same considerations apply when designing connected systems in which sensors, edge devices, cloud services and operational applications must work together.
Four IoT service interaction patterns
In the monitoring scenario, a wearable sensor supplies heart-rate history, a smartwatch analyzes readings, a cloud service identifies the nearest ambulance, and the ambulance provides assistance. The workflow is a useful way to compare how services can cooperate toward one outcome.
Direct interactions
In a direct interaction, a sending service calls a receiving service without a mediator. This commonly uses request-response communication, such as an RPC or REST API call. The sender initiates the control flow, and the receiver returns a response. In the example, the analysis service requests the latest sensor reading and, if necessary, calls the emergency service.
This approach makes the sequence of calls relatively clear, but services need to know how to reach their receivers. If a receiver moves or its location changes, the sender may need an updated reference or a discovery mechanism. Direct coupling can also make changes to workflow logic affect the services that contain it.
Indirect interactions through a service bus
An indirect interaction inserts a service bus between senders and receivers. Services register their interfaces with the bus, which routes requests and responses. The intermediary provides location transparency to participating services: they communicate through the bus rather than needing to address each other directly.
The trade-off is that the bus becomes a central point through which interactions pass. This can simplify routing and application design, but it adds an intermediary and an additional network hop. A bus used only for routing differs from one that also contains workflow or business coordination logic; the latter begins to take on an orchestration role.
Event-driven interactions
Event-driven services use a publish-subscribe model. Producers publish events, and interested consumers receive and act on them. Communication is often asynchronous and follows a “fire and forget” pattern: a producer does not necessarily receive a response from each consumer.
Events can be exchanged peer to peer, with each service managing its own queue, or through a broker or event bus. In the monitoring scenario, the analysis service could consume sensor-reading events, publish an emergency event when needed, and receive an event reporting assistance status. This reduces the need for publishers and subscribers to know one another’s locations. However, the workflow is implicit in the event exchanges, which can make it harder to trace execution, debug behavior or see the complete control path.
Exogenous interactions and orchestration
With exogenous interactions, a coordinator defines the workflow while services provide functionality. The services do not determine the sequence by calling one another; instead, a coordinator passes control among them. A single workflow engine is one form of orchestration. In multi-level arrangements, multiple coordinators divide and coordinate work across a hierarchy.
Separating workflow logic from service functionality can make services more reusable when application requirements change. A new step, for example, can be introduced by changing coordination logic rather than embedding the change in several service implementations. Multi-level coordination can also distribute control among multiple coordinators rather than relying on one central workflow engine.
How interaction patterns affect IoT scalability
The analysis assesses each schema using four architectural properties. These are useful questions for engineering teams evaluating a connected system:
- Is control flow explicit? A visible workflow can make execution easier to monitor and maintain. Direct and indirect interactions place control in service logic, while exogenous interactions make it explicit in coordinators. Event-driven control is implicit in the exchange of events.
- Are control and computation separate? When workflow decisions live inside service code, a change to the application sequence may require changes to the services themselves. Exogenous interactions separate coordinator-defined control from the functionality provided by services.
- Is control decentralized? Direct interactions, peer-to-peer event exchanges and multi-level exogenous interactions distribute control across more than one entity. A bus, broker or single workflow engine introduces a central intermediary or coordinator.
- Do services have location transparency? Indirect, event-driven and exogenous interactions can hide the location of other services from participants. Direct interactions generally require a sender to know the receiver’s location unless discovery is added.
In this qualitative comparison, multi-level exogenous interactions are the only schema assessed as meeting all four criteria. One-level orchestration meets three, while direct, indirect and peer-to-peer event-driven interactions meet two each. Broker-based event-driven interactions meet one. These results describe architectural characteristics, not a universal performance ranking: the analysis does not quantitatively measure throughput or establish that one design is best for every application.
Applying the comparison to industrial automation
For automation engineers and system integrators, the practical question is which trade-offs matter most in a particular deployment. A simple direct call may suit a stable, well-understood connection. A brokered or event-based pattern may be more appropriate when services should communicate asynchronously or without knowing each other’s locations. Where the workflow must be visible and service logic should remain independent of changing process coordination, orchestration may be worth considering.
Central coordination can simplify workflow design, but it can also place more communication through a single point. Decentralized designs avoid some intermediary dependencies, yet their control paths may be harder to visualize or manage. Designs can also combine patterns: for example, events can support asynchronous exchanges while a direct interaction handles a specific request-response step. The choice should reflect operational needs, network conditions, device constraints and the required approach to monitoring and change.
Conclusion: design the interaction model deliberately
IoT scalability depends not only on how many devices a system can connect, but also on how its services coordinate as the application grows and changes. Direct calls, service buses, publish-subscribe messaging and orchestration each shape control visibility, coupling, decentralization and location awareness differently. Teams should assess these properties alongside network performance and implementation constraints. Making the interaction model explicit early can help create connected systems that are easier to operate, adapt and maintain.
Access Official Technical PDF