Timestream Data Design & Ingestion

After the first introduction to AWS Timestream, let's get into it.

Let's first spin some data. As an example, let's assume that you have an IoT device that will send you the following event every so often:

{
    "event_id": 123456,
    "created_time": 1604355422089,
    "device_id": "test_device",
    "usecase": "retail",
    "device_battery": 80,
    "temperature": 60.34,
    "weight": 20
}

`

So how do you get this into AWS Timestream?

As with every database, you start with your schema.
AWS Timestream splits its data into measures and dimensions. How should we break our sample event into these categories?
As the docs say, fields that "describe" the data should be the dimensions. In the event above, I would see the event_id, device_id, and usecase as the dimensions. Conversely device_battery, temperature and, weight would be the measures. And created_time as the time dimension of course.

The weird thing about AWS Timestream is that you will add this data as 3 rows, as it contains three measurements:

event_id	device_id	usecase	measure_value::BIGINT	measure_value::DOUBLE	measure_name	time
123456	test_device	retail	80		device_battery	1604355422089
123456	test_device	retail		60.34	temperature	1604355422089
123456	test_device	retail	20		weight	1604355422089