Multiverse::Graviton

Real time geospatial data analytic & visualization platform

Background

Booming of geolocation data with high volume, variety, & velocity.
 Example: 


• Gojek data: driver real time location, order location, merchant location, etc.
• iOT data: humidity sensor, air pollution sensor, river level sensor, etc

Problem

Business perspective:

The need to visualise & analyse real time geospatial data.

Technical perspective:

The need of platform, library, data modelling/architecture that can easily visualise & analyse realtime geospatial data.

Challenges

Realtime Geospatial data characteristics:

• high volume
• high variety
• high velocity

Some people call it: BIG DATA

Multiverse solution

In this project we provide:

• Data modelling (convention & configuration)
• Data architecture
• Data visualisation

...that can be reusable for any kind of time series geospatial/geolocation data.

Our goal is to create architecture & data modelling that:

• Optimised for timeseries realtime geolocation data.
• Horizontally scalable
• Fault tolerant
• Produce near realtime insight

Current available query of data aggregation:

• Query minutely aggregated geolocation data per event_name.
• Query hourly aggregated geolocation data per event_name.
• Query daily aggregated geolocation data per event_name.

Guideline

Event data format:

   {
      "header": {
        "event_name": "event_name",
        "timestamp": utc_unix_formatted_timestamp
      },
      "body": {
        "location": [lat, long],
        "other_data1": "sample1",
        "other_data2": "sample2",
      }
    }

Example:

   {
      "header": {
        "event_name": "gofood.order.canceled",
        "timestamp": 1490492075
      },
      "body": {
        "location": [-6.178005,106.7881563],
        "order_id": "order-1234",
        "customer_id": "customer-345",
        "merchant_id": "merchant-678",
        "reason": "waited.too.long"
      }
    }

Architecture

1. Apache Kafka

   High-throughput, low-latency platform for handling real-time high load of Gojek event data.

2. Apache Storm

   Distributed realtime computation system to compute/aggregate data stream from Kafka into analytical data model.

3. Aerospike

   NoSQL database for real-time operational applications to save analytical data.

4. Rails

   To build API of analytical data (query abstraction) from Aerospike

5. d3.js

   To build rich interactive front end visualization of analytical data.

Sample front-end screenshot

heatmap of order.canceled event today

Team:

Multiverse

• Andri Setiawan
• Azhar Amir
• Fauzan Qadri