Cluster Operation

StreamPipes' microservice approach includes the UI, the StreamPipes core for pipeline management, and all extensions, such as Connect adapters and pipeline elements. Flexible orchestration is available using Docker. In addition to the widely used AMD-based architectures, StreamPipes now also supports ARM-based systems.

The ARM support means that, for certain use cases, individual algorithm containers can be started on small edge devices, such as a Jetson Nano or Raspberry Pi, while the pipeline management core is hosted centrally. This is achieved by means of multi-architecture Docker images available on Docker Hub. These images are annotated via the Docker Manifest feature, so the user does not need to adjust the image tags in deployment descriptions. With a combination of architecture-specific image tags and an associated Docker manifest, you can create a one-size-fits-all image description that agnostically retrieves the right image of Docker Hub for the system architecture.

Experience has shown that containerizing services makes it possible to implement different operation options, from single server instances to full cluster operations. For single-server deployment, the StreamPipes environment can be quickly and easily launched using Docker Compose, a tool in the Docker ecosystem for defining and running multi-container Docker applications. The StreamPipes services defined in a YAML file are configured this way and then started locally with a single command.

Especially in use cases where high-powered computing resources are not available internally, or where the cloud is not an option, even a user without in-depth Docker skills can set up an executable StreamPipes instance in a few minutes. In addition to server operation, you can also provision small, portable mini PCs with StreamPipes.

It is also possible to use StreamPipes in distributed clusters. For this purpose, you can operate the individual microservices in a Kubernetes infrastructure, using Kubernetes' Helm package manger to reduce complexity. Helm lets you combine relatively complex Kubernetes YAML manifests into a single package. You can install StreamPipes' own Helm chart very easily in a Kubernetes cluster using a one-liner:

$ cd incubator-streampipes-installer/k8s
$ helm install streampipes ./

This also means that you can create Kubernetes clusters of edge nodes on the shop floor, as well as on centralized back-end servers. StreamPipes Connect can then connect the data at an early stage directly at the machine and, if necessary, set up processing algorithms for transformation, filtering, enrichment, and so on. This approach ensures that you don't necessarily have to transmit all the raw data, which is often not feasible due to restrictions such as latency, available bandwidth, or data sovereignty.

A blog post on the StreamPipes website contains detailed information about using the StreamPipes Helm chart in an example Raspberry Pi 4 Kubernetes cluster based on Rancher's lightweight K3s distribution.

Conclusions

The relatively young Apache StreamPipes incubator project by the Apache Software Foundation seeks to improve the accessibility of data stream-based applications for business users. With the underlying microservices approach, StreamPipes seeks to achieve the greatest possible reusability of the individual components. In the end, however, you'll need to decide whether the flexibility benefits of a modular solution exceed the benefits of a customized, programmed application.

In addition to StreamPipes and the popular Apache Flink and Apache Kafka tools, the Apache Software Foundation offers other projects that are useful in IoT deployment scenarios. The top-level Apache PLC4X project, for example, focuses especially on connecting machine data in an industrial context. Apache IoTDB is a relatively new database that specializes in persisting time series. The Apache Software Foundation maintains a strong community-driven approach to development. The developer community welcomes contributions of all kinds, enabling everyone to contribute to building a strong, open source IoT ecosystem.