Get fresh updates from Hortonworks by email

Once a month, receive latest insights, trends, analytics information and knowledge of Big Data.


Get Started


Ready to Get Started?

Download sandbox

How can we help you?

closeClose button
HDF > Develop with Hadoop > Real World Examples

Analyze IoT Weather Station Data via Connected Data Architecture

IoT and Connected Data Architecture Concepts

cloud Ready to Get Started?



Anytime there are at least two platforms connected: HDF and HDP, that environment is called Connected Data Architecture. For this tutorial, we are building a Connected Data Architecture that incorporates: MiNiFi agent based at the Raspberry Pi, HDF docker sandbox and HDP docker sandbox both located within the docker engine. The purpose of the concepts section is to dive into each tool that is used within this tutorial.

Apache NiFi is a dataflow management framework that makes it possible to ingest data from any data source and transport it to any destination, so that developers can focus on data processing implementation. We will use Apache MiNiFi to ingest the sensor readings from the R-Pi and transport that data to NiFi to send to HBase. HBase is a noSQL database and we will use it because it’s efficient at storing unstructured data. We will use Apache Phoenix to map to the HBase table and Perform SQL queries against the HBase table. Apache Zeppelin will be the business reporting tool to visualize these queries we perform against the HBase table.

In the concepts tutorial, the goal is to give you information on the background of each hardware/software tool used to build this Connected Data Architecture System for use cases, such as the IoT Weather Station.


  • None


  • Raspberry Pi
  • Sense HAT Sensor Functionality
  • Docker
  • HDF Sandbox Docker Container
  • Apache NiFi
  • Apache MiNiFi
  • HDP Sandbox Docker Container
  • Apache Zookeeper
  • Apache HBase
  • Apache Phoenix
  • Apache Zeppelin
  • Connected Data Architecture and IoT Weather Station
  • Summary
  • Further Reading

Raspberry Pi

What is a Raspberry Pi?

The Raspberry Pi (R-Pi) 3 is a microprocessor or computer with an open-source platform commonly used for beginners learning to code to practitioners building Internet of Things (IoT) related Applications. This embedded device has a 1.2 GHz ARMv8 CPU, 1GB of memory, integrated Wi-Fi and Bluetooth. The R-Pi comes with various General Purpose Input Output (GPIO) pins, input/output ports for connecting the device to external peripherals, such as sensors, keyboards, mouses and other peripherals. As can be seen in Figure 1, the R-Pi is connected to the internet via Ethernet port, a monitor via HDMI port, keyboard and mouse via USB port and powered by 12V power supply. This device has the capability to run various operating systems, such as Linux. Additionally, it can run an instance of Apache NiFi and for embedded devices that have a limited amount of memory, we can run Apache MiNiFi.


Figure 1: Raspberry Pi

Internet of Things on R-Pi

The R-Pi is not just a platform for building IoT projects, it is a super platform for learning about IoT. R-Pi is a great way to gain practical experience with IoT. According to IBM Watson Internet of Things, the R-Pi IoT platform can be used for the factory, environment, sports, vehicles, buildings, home and retail. All these R-Pi IoT platform use cases have in common that data is processed, which can result in augmented productivity in factories, enhanced environmental stewardship initiatives, provided winning strategies in sports, enhanced driving experience, better decision making, enhanced resident safety and security and customized and improved shopping experience in retail.

Sense HAT Functionality


Figure 2: Sense HAT

The Sense HAT comes with an 8×8 LED Matrix, five-button joystick and the following sensors: gyroscope, accelerometer, magnetometer, temperature, barometric pressure and humidity.

What exactly does the Sense HAT Sensor Measure?

The Sense HAT sensors enable users to measure orientation via an 3D accelerometer, 3D gyroscope and 3D magnetometer combined into one chip LSM9DS1. The Sense HAT also functions to measure air pressure and temperature via barometric pressure and temperature combined into the LPS25H chip. The HAT can monitor the percentage of humidity in correlation with temperature in the air via humidity and temperature sensor HTS221. All three of the these sensors are I2C.

How does the Sense HAT Sensor Transfer Data to the R-Pi?

The Sense HAT sensor uses I2C, a communication protocol, to transfer data to the R-Pi and other devices. I2C requires two shared lines: serial clock signal (SCL) and bidirectional data transfers (SDA). Every I2C device uses a 7-bit address, which allows for more than 120 devices sharing the bus, and freely communicate with them one at a time on as-needed basis.

What is an Advantage of I2C Sensors?

I2C makes it possible to have multiple devices in connection with the R-Pi, each having a unique address and can be set by updating the settings on the Pi. It will be easier to verify everything is working because one can see all the devices connected to the Pi.


What is Docker?

Docker is a an open source platform for developers and sysadmins to develop, ship, and run applications. Docker provides faster delivery of applications, allows users to deploy and scale easily, which results in easier maintenance. This platform includes a Docker Engine, which is a lightweight and powerful open source containerization technology. This technology incorporates work flow for building and containerizing your applications. Docker containerization incorporates Docker images and containers. Docker images contain applications or services that can easily deploy an application into a testing, staging and production environment known as containers. For instance, with one Docker image, you can deploy multiple containers of that Docker image instance. Containers are a way to package and run an application, such as a framework: Hadoop, Spark, NiFi in an isolated environment. Containers are different from virtual machine because they do not need the extra layer of a hypervisor, instead they run directly on the host machine’s kernel. Users are able to share their docker images and containers through the Docker Hub.

Docker Architecture


Figure 3: Docker Architecture

In the docker architecture above, Docker registry are services used for storing Docker images, such as Docker Hub. Docker Host is the computer Docker runs on. Diving deeper into the host, you can see the Docker Daemon, which is used to create and manage Docker objects, such as images, containers, networks and volumes. The user or client is able to interact with Docker daemon via Client Docker CLI. Additionally, the CLI are scripts or direct commands entered by the user. The Docker daemon is a long-running program also known as a server. The CLI utilizes Docker’s REST API to interact with the Docker daemon. As you can observe, the Docker Engine is a client-server application comprised of Client Docker CLI, REST API and Docker daemon.

You will use Docker as the backbone to deploy the Connected Data Architecture: IoT Devices, HDF Sandbox Container and HDP Sandbox Container.


Figure 4: Connected Data Architecture

In the Connected Data Architecture, the Sense HAT and Raspberry Pi will be located in the Internet of Anything (alias Internet of Things), HDF Sandbox Container and HDP Sandbox Container will run in their own containers connected in a simulated network by Docker’s default network feature known as Bridge.

HDF Sandbox Docker Container

HDF Sandbox comes in different flavors: VirtualBox, VMware and Docker. You will be using the HDF Sandbox Docker container to build this IoT Weather Station via Connected Data Architecture. Hortonworks DataFlow (HDF) Sandbox is a way to deploy HDF into an isolated environment for testing, staging and sometimes production. HDF is a application stack framework used to process data-in-motion. HDF comes with frameworks: Zookeeper, Storm, Ambari Infra, Ambari Metrics, Kafka, Log Search, Ranger and NiFi.

Apache NiFi

NiFi is a robust and secure framework for ingesting data from various sources, performing simple transformations on that data and transporting it across a multitude of systems. The NiFi UI provides flexibility to allow teams to simultaneously change flows on the same machine. NiFi uses SAN or RAID storage for the data it ingests and the provenance data manipulation events it generates. Provenance is a record of events in the NiFi UI that shows how data manipulation occurs while data flows throughout the components, known as processors, in the NiFi flow. NiFi, program sized at approximately 800MB, has over 190 processors for custom integration with various systems and operations.

Apache MiNiFi

MiNiFi was built to live on the edge for ingesting data at the central location of where it is born, then transport that data to your data center where NiFi lives. MiNiFi comes in two flavors Java or C++ agent. These agents access data from IoT and Desktop level devices. The idea behind MiNiFi is to be able to get as close to the data as possible from any particular location no matter how small the footprint on a particular embedded device.

Visualization of MiNiFi and NiFi Place in IoT


Figure 5: MiNiFi to NiFi

HDP Sandbox Docker Container

HDP Sandbox comes in different flavors: VirtualBox, VMware and Docker. You will be using the HDP Sandbox Docker container to build this IoT Weather Station via Connected Data Architecture. Hortonworks Data Platform (HDP) Sandbox is a way to deploy HDP into an isolated environment for testing, staging and sometimes production. HDP is a application stack framework used to process data-at-rest. HDP comes with various frameworks: HDFS, Yarn + MapReduce2, HBase, Phoenix, Zeppelin, etc.

Apache Zookeeper


Zookeeper is an open-source coordination service designed for distributed applications. It uses a data model similar to the tree structure of file systems. With these apps, a set of primitives are exposed to implement higher level services for cases of synchronization, configuration maintenance, groups and naming. Typically building coordination services are difficult and prone to errors like race conditions and deadlocks. The goal with Zookeeper is to prevent developers from having to build coordination services from scratch across distributed applications.

Design Goals

Zookeeper is designed to be simple, replicated, ordered and fast. Ideal read/write ratio is close to 10:1 resulting in fast execution against workloads. Transaction records are maintained and can be incorporated for higher-level abstractions. To reflect the order of all transactions, every update is stamped with a number.

How the Zookeeper Service Works

In a Zookeeper cluster, you will usually have servers or nodes that make up the Zookeeper service that all know each other exist. They hold in-memory image of the state, transaction logs and snapshots in the persistent store. The Zookeeper service remains available as long as the servers are up and running. The process between client and Zookeeper involves the client making a connection to a Zookeeper server. The client maintains the TCP connection and sends requests, heart beats along with obtaining responses and watch events. Once Zookeeper crashes, clients will connect to another server.

Apache HBase

Apache HBase is a noSQL database programmed in Java, but unlike other noSQL databases, it provides strong data consistency on reads and writes. HBase is a column-oriented key/value data store implemented to run on top of Hadoop Distributed File System (HDFS). HBase scales out horizontally in distributed compute clusters and supports rapid table-update rates. HBase focuses on scale, which enables it to handle very large database tables. Common scenarios include HBase tables that hold billions of rows and millions of columns. Another example is Facebook utilizes HBase as a structured data handler for its messaging infrastructure.

Critical part of the HBase architecture is utilization of the master nodes to manage region servers that distribute and process parts of data tables. HBase is part of the Hadoop ecosystem along with other services such as Zookeeper, Phoenix, Zeppelin.

Apache Phoenix

Apache Phoenix provides the flexibility of late-bound, schema-on-read capabilities from NoSQL technology by leveraging HBase as its backing store. Phoenix has the power of standard SQL and JDBC APIs with full ACID transaction capabilities. Phoenix also enables online transaction processing (OLTP) and operational analytics in Hadoop specifically for low latency applications. Phoenix comes fully integrated to work with other products in the Hadoop ecosystem, such as Spark and Hive.

Apache Zeppelin

Apache Zeppelin is a data science notebook that allows users to use interpreters to visualize their data with line, bar, pie, scatter charts and various other visualizations. One particular interpreter we will utilize is Phoenix that way we can visualize our weather data.

Connected Data Architecture and IoT Weather Station

We have talked about each component within the project individually, now we will discuss at high level how each component connects to each other.


Figure 6: IoT Weather Station and Connected Data Architecture Integration

IoT Weather Station at the Raspberry Pi + Sense HAT + MiNiFi

The IoT Weather Station is based at the Raspberry Pi and Sense HAT. At this location, the Raspberry Pi executes a python script that controls the Sense HAT sensors to generate weather data. MiNiFi is agent software that resides on the Raspberry Pi and acts as an edge node that ingests the raw data from the sensor.

Connected Data Architecture at HDF + HDP Sandbox Nodes

The combination of HDF and HDP create the idea of Connected Data Architecture. For prototyping purposes, you will be running a single node HDF sandbox container and a single node HDP sandbox container inside a docker containerized network. The way that sensor data gets into the Connected Data Architecture is through MiNiFi’s connection to HDF NiFi via Site-To-Site. Site-To-Site is a protocol that allows external agents to connect to NiFi. NiFi preprocesses the sensor data and adds geographic enrichment. NiFi located on the HDF Sandbox Container then makes a connection to HBase via NiFi’s HBase Client Service Controller. This Controller Service enables NiFi to connect to a remote HBase located on the HDP Sandbox Container. The way that NiFi is able to find the location of HBase is through communicating with the Zookeeper client. With successful communication from Zookeeper, NiFi is able to locate the location of HBase and store data in an HBaes table. Once the data is stored in HBase, Phoenix is used to map to the HBase table. Visualization is performed using Zeppelin’s Phoenix interpreter.

Note: “Single node” just means that you have one computer or docker container that has an HDF stack or HDP stack installed on it. On the other hand, a multinode could be HDF stack installed across multiple computers or docker containers.


Congratulations, you’ve finished the concepts tutorial! Now you are familiar with the technologies you will be utilizing in the tutorial series and will have a better understanding of each tools purpose in deploying this IoT Weather Station by way of Connected Data Architecture.

Further Reading