Sample IoT application simulating a monitoring dashboard for a fleet management/shipping company.
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 27 commits ahead of baghelamit:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

IoT Fleet Management

YugaByte DB is world's 1st open source database that is both NoSQL (Cassandra & Redis compatible) and SQL (PostgreSQL compatible) at the same time. It is purpose-built to power fast-growing online services on public, private and hybrid clouds with transactional data integrity, low latency, high throughput and multi-region scalability while also using popular NoSQL and SQL APIs.

This is a sample application that shows how real-time streaming applications (such as those in the IoT vertical) can leverage YugaByte DB as a highly reliable, elastic operational database. It uses YugaByte DB's Cassandra-compatible YCQL API.


Here is a brief description of the scenario.

Assume that a fleet management company wants to track their fleet of vehicles, which are of different types (18 Wheelers, busses, large trucks, etc).

Below is a view of the dashboard of the running app.

IoT Fleet Management Dashboard

The above dashboard can be used to monitor the different vehicle types and the routes they have taken both over the lifetime of the app as well as over the last 30 second window. It also points out the trucks that are near road closures, which might cause a delay in the shipping schedule.


The IoT Fleet Management application contains the following four components:

  • IoT Kafka Producer This component emulates data being emitted from a connected vehicle, and generates data for the Kafka topic iot-data-event. The data emitted is of the format shown below.

    {"vehicleId":"0bf45cac-d1b8-4364-a906-980e1c2bdbcb","vehicleType":"Taxi","routeId":"Route-37","longitude":"-95.255615","latitude":"33.49808","timestamp":"2017-10-16 12:31:03","speed":49.0,"fuelLevel":38.0}
  • IoT Real-Time Data Processor This component reads data from Kafka topic iot-data-event and computes the following:

    • Total traffic snapshot
    • Last 30 seconds traffic snapshot
    • Vehicles near a point of interest

    There are two ways the app can perform this analysis. First is through KSQL, Confluent's SQL-like streaming query language for Kafka, and second is through Apache Spark as an external stream processing engine.

  • IoT Database This component is based on YugaByte DB. YugaByte DB's Cassandra-compatible YCQL API is used to integrate with other components of the app.

  • IoT Spring Boot Dashboard This app uses the Java Spring Boot framework with its integration for Cassandra as the data layer, using the Cassandra Query Language (CQL) internally.

Architecture with KSQL

Architecture with KSQL

Architecture with Apache Spark Streaming

Architecture with Apache Spark Streaming


For building these projects it requires following tools. Please refer files of individual projects for more details.

  • JDK - 1.8 +
  • Maven - 3.3 +
  • Confluent Open Source - 5.0.0 (we assume this is installed in the ~/yb-kafka/confluent-os/confluent-5.0.0 directory).
  • YugaByte Connect sink - 1.0.0 (clone this into ~/yb-kafka/yb-kafka-connector).

Steps to setup the environment

  1. Build the required binaries.

    mvn package
  2. Download Confluent Open Source from This is a manual step, since an email id is needed to register (as of Nov 2018). Unbundle the content of the tar.gz to location ~/yb-kafka/confluent-os/confluent-5.0.0 using these steps.

    mkdir -p ~/yb-kafka/confluent-os
    cd ~/yb-kafka/confluent-os
    tar -xvf confluent-5.0.0-2.11.tar.gz
  3. Include dependent components into Kafka connectors:

  • Build the jar from this repo and copy it for use by Kafka:

    cd  ~/yb-kafka/
    git clone
    cd  ~/yb-kafka/yb-kafka-connector/
    mvn clean install -DskipTests
    mkdir ~/yb-kafka/confluent-os/confluent-5.0.0/share/java/kafka-connect-yugabyte/
    cp  ~/yb-kafka/yb-kafka-connector/target/yb-kafka-connnector-1.0.0.jar ~/yb-kafka/confluent-os/confluent-5.0.0/share/java/kafka-connect-yugabyte/
  • Setup the property files for use by Connect Sink.

    cp iot-ksql-processor/resources/kafka.* ~/yb-kafka/confluent-os/confluent-5.0.0/etc/kafka/
    cp iot-ksql-processor/resources/* ~/yb-kafka/confluent-os/confluent-5.0.0/etc/kafka-connect-yugabyte
  • Download the dependent jars from maven central repository using the following commands.

    cd ~/yb-kafka/confluent-os/confluent-5.0.0/share/java/kafka-connect-yugabyte/

    The final list of jars should look like this:

     $ ls -al
      -rw-r--r--@    85449 Oct 27  2013 metrics-core-3.0.1.jar
      -rw-r--r--@  3823147 Oct 27 15:18 netty-all-4.1.25.Final.jar
      -rw-r--r--   1100520 Oct 29 11:18 cassandra-driver-core-3.2.0-yb-18.jar
      -rw-r--r--     14934 Oct 29 11:19 yb-kafka-connnector-1.0.0.jar
  1. Do the following to run Kafka and related components:

    export PATH=$PATH:~/yb-kafka/confluent-os/confluent-5.0.0/bin
    confluent start ksql-server
    confluent status

    The output for the confluent status should look like

    control-center is [DOWN]
    ksql-server is [UP]
    connect is [DOWN]
    kafka-rest is [DOWN]
    schema-registry is [UP]
    kafka is [UP]
    zookeeper is [UP]

    Note: It is required that the DOWN components in this list are not actually enabled.

  2. Create the origin Kafka topic

     ~/yb-kafka/confluent-os/confluent-5.0.0/bin/kafka-topics --create --zookeeper localhost:2181 --replication-factor 1 --partitions 1 --topic iot-data-event

    Note: This is needed to be done only the first time.

  3. Install YugaByte DB.

  4. Create the YugaByte DB tables

    • Create the keyspaces and tables by running the following command. You can find cqlsh in the bin sub-directory located inside the YugaByte installation folder.
      $> cqlsh -f resources/IoTData.cql
  5. Run the origin topic YugaByte DB Connect Sink

    cd ~/yb-kafka/confluent-os/confluent-5.0.0
    nohup ./bin/connect-standalone ./etc/kafka/ ./etc/kafka-connect-yugabyte/ >& origin_sink.txt &

    This will insert the origin topic data into the YugaByte DB CQL table TrafficKeySpace.Origin_Table.

Running the application

From the top level directory of this repo, run the following

  1. Start the data producer.

    java -jar iot-kafka-producer/target/iot-kafka-producer-1.0.0.jar

    It should start emitting data points to the Kafka topic. You should see something like the following as the output on the console:

    2017-10-16 12:31:52 INFO  IoTDataEncoder:28 - {"vehicleId":"0bf45cac-d1b8-4364-a906-980e1c2bdbcb","vehicleType":"Taxi","routeId":"Route-37","longitude":"-95.255615","latitude":"33.49808","timestamp":"2017-10-16 12:31:03","speed":49.0,"fuelLevel":38.0}
    2017-10-16 12:31:53 INFO  IoTDataEncoder:28 - {"vehicleId":"600863bc-c918-4c8e-a90b-7d66db4958e0","vehicleType":"18 Wheeler","routeId":"Route-43","longitude":"-97.918175","latitude":"35.78791","timestamp":"2017-10-16 12:31:03","speed":59.0,"fuelLevel":12.0}
  2. Start the data processing application Use either of these options:

  • Spark
    • Run the spark app using this
      java -jar iot-spark-processor/target/iot-spark-processor-1.0.0.jar
  • KSQL
    • Setup the KSQL tables/streams
      ksql <<EOF
      RUN SCRIPT './iot-ksql-processor/setup_streams.ksql';
    • Run the connect sink from KSQL processed data
      cd ~/yb-kafka/confluent-os/confluent-5.0.0
      nohup ./bin/connect-standalone ./etc/kafka/ ./etc/kafka-connect-yugabyte/ ./etc/kafka-connect-yugabyte/ ./etc/kafka-connect-yugabyte/ >& ksql_sink.txt &
  1. Start the UI application.

    java -jar iot-springboot-dashboard/target/iot-springboot-dashboard-1.0.0.jar
  2. Now open the dashboard UI in a web browser. The application will refresh itself periodically.