HTM-MoClu

Hierarchical Temporal Memory Models Cluster implementation

What is Htm-Moclu?

Short for Hierarchical Temporal Memory Models Cluster. Numenta presented HtmEngine which is a set of long-running services upon which a scalable, real-time, and production-ready scalar anomaly detection application may be built. As-is, HtmEngine is only configured to be scaled vertically. Htm-Moclu provides a similar platform for htm.java applications, and has the ability to scale horizontally using multiple servers.

You can have several Web servers and several servers just to handle model data all of them working as a Cluster.

Architecture

Models, anomaly detection, and prediction provided by htm.java
Cluster implemented using Akka Cluster + Sharing + Persistence
Data store: Mongodb
Rest API: Lift Framework (moclu-http)

Installation

Requirements

SBT

MongoDB

After installing MongoDB you will need to start mongod service and use default settings, otherwise you will need to configure application.conf

Starting the cluster

git clone https://github.com/antidata/htm-moclu.git
cd htm-moclu
sbt publishLocal
sbt console

Starting the Web server

cd htm-moclu/moclu-http
sbt ~container:start

Using the REST API

####Create model - Endpoint ####

/create/{model id}

Example

/create/24StreetSensor

Request Payload

{ }

Response

{
  "status":200,
  "msg":"Htm Model 24StreetSensor created"
}

####Send data - Endpoint #### /event/{model id}

Example

/event/24StreetSensor

Request Payload

{ 
  "value":12.2,
  "timestamp":"7/2/10 1:11" 
}

Response

{
  "status":200,
  "msg":"Applied event Htm Model 24StreetSensor",
  "data":{
    "id":"24StreetSensor",
    "anomalyScore":1.0,
    "prediction":"12.2"
  }
}

####Get submitted data - Endpoint #### /getData/{model id}

Example

/getData/24StreetSensor

Request Payload

{ }

Response

{
  "status":200,
  "data":[{
    "value":12.2,
    "timestamp":1278051060000,
    "anomalyScore":1.0
  },{
    "value":13.2,
    "timestamp":1278051120000,
    "anomalyScore":1.0
  },{
    "value":14.2,
    "timestamp":1278051180000,
    "anomalyScore":1.0
  },{
    "value":12.2,
    "timestamp":1278051300000,
    "anomalyScore":1.0
  },{
    "value":13.2,
    "timestamp":1278051360000,
    "anomalyScore":0.0
  }]
}

Few examples using wget

Create

wget -O- --post-data='{}' --header=Content-Type:application/json "http://localhost:8080/create/24StreetSensor"

Send event data

wget -O- --post-data='{"value":13.2,"timestamp":"7/2/10 1:16"}' --header=Content-Type:application/json "http://localhost:8080/event/24StreetSensor"

Get events data

wget -O- --post-data='{}' --header=Content-Type:application/json "http://localhost:8080/getData/24StreetSensor"

Setting up the Cluster

Configuring MongoDB

To set the IP address where you have MongoDB running just add the following parameter to SBT:

-DMONGOHOST={IP}

For example having Mongo running on 192.168.1.144 the config will look like

-DMONGOHOST=192.168.1.144

This configuration is the same for all nodes.

Cluster Seeds

You can have one or more seeds for your Cluster, you can start trying with just one.

add the following parameter to SBT:

-DSEEDHOST={IP}

In the case you have the Seed running on 192.168.1.122, after modifying:

-DSEEDHOST=192.168.1.122

If you want to add more seed nodes you will need to update the application.conf and add more parameters.

Local IP

Finally we need to specify the local IP with the following parameter

-DHOST=192.168.1.123

In case you are starting the node in 192.168.1.123

Putting all together

$ sbt -Djava.library.path=./sigar -DHOST=192.168.1.123 -DSEEDHOST=192.168.1.122 -DMONGOHOST=192.168.1.144 -Xmx8096M -Xss2M