MIST: High-performance IoT Stream Processing

MIST is a stream processing system that is optimized to handle large numbers of IoT stream queries. MIST is built on top of Apache REEF.

Requirements

• Java 1.8
• Maven

How to build and run MIST

1. Build MIST:

   git clone https://github.com/snuspl/mist
   cd mist
   Set $MIST_HOME (export MIST_HOME=`pwd`)
   mvn clean install
   

2. Run MIST:

   ./bin/start.sh
   [-? Print help]
   -num_threads The number of MIST threads
   -num_task_cores The number of cores of MIST task
   [-num_tasks The number of mist tasks]
   [-port The port number of the RPC server that communicates with MIST client]
   [-runtime Type of the MIST Driver runtime (yarn or local. The default value is local runtime)]
   -task_mem_size The size of MIST task memory (MB)
   

MIST examples

HelloMist

1. Description

   The example shows how to implement a simple stateless query and submit it to the MIST. The code sets the environment(Hostnames and ports of MIST, source, sink), generates a simple stateless query, and submits it to the MIST. The query reads strings from a source server, filter strings which start with "HelloMIST:", trim "HelloMIST:" part of the filtered strings, and send them to a sink server.

2. Run HelloMist

   # 0. Build MIST first!(See above)
   # 1. Run MIST
   ./bin/start.sh -num_threads 1 -num_task_cores 1 -task_mem_size 1024
   # 2. Launch source server (You can simply use netcat)
    nc -lk 20331
   # 3. Run a script for HelloMist
   ./bin/run_example.sh HelloMist
   [-? Print help]
   [-d Address of running MIST driver in the form of hostname:port]
   [-s Address of running source server in the form of hostname:port]
   
   # 4. Publish data stream 
    nc -lk 20331
    HelloMIST: hello!  (Then HelloMist query will filter this string and print "hello!" message to the console) 
   

Multi-query submission

To submit stream queries, application developers first submit their jar files to MIST (The HelloMist example automatically submits the jar file and query to MIST).

# submit jar file
./bin/submit_jar.sh ./mist-examples/target/mist-examples-0.2-SNAPSHOT.jar

Then, it will return the app identifier

App identifier: 0

We can use this app identifier to submit multiple stream queries of the app. We will use EMQ as an MQTT message broker. Please download the EMQ and start it.

emqttd start

The default port of the emqttd that subscribes and publishes data is 1883. We then submit a stream query that filters data according to the user-defined prefix. This query subscribes/publishes the data stream from/to the EMQ.

./bin/run_example.sh MqttFilterApplication -source_topic /src/1 -sink_topic /sink/1 -filtered_string "HelloMIST:" -app_id 0

Then, the query will subscribe data from /src/1 EMQ topic, filter data that contains HelloMIST: as a prefix, and publishes the data to /sink/1 EMQ topic.

To publish and subscribe the topic, we can use mosquitto_pub and mosquitto_sub. We can subscribe the /sink/1 topic by:

mosquitto_sub -h 127.0.0.1 -t /sink/1

We can publish the /src/1 topic by :

mosquitto_pub -h 127.0.0.1 -t /src/1 -m "HelloMIST: hello!"

We can submit another stream query that filters HelloWorld:

./bin/run_example.sh MqttFilterApplication -source_topic /src/2 -sink_topic /sink/2 -filtered_string "HelloWorld:" -app_id 0