Amazon Timestreamでユーザー定義のパーティションを試してみた

AWSTemplateFormatVersion: "2010-09-09"
Description: Amazon Timestream Many Data Sample use Partition Key

Resources:
  TimestreamDatabase:
    Type: AWS::Timestream::Database
    Properties:
      DatabaseName: ManyDataSampleUsePartitionKeyDatabase

  TimestreamTable:
    Type: AWS::Timestream::Table
    Properties:
      DatabaseName: !Ref TimestreamDatabase
      TableName: many-data-sample-use-partition-key-table
      Schema:
        CompositePartitionKey:
         - Name: deviceId
           Type: DIMENSION
           EnforcementInRecord: REQUIRED
      RetentionProperties:
        MemoryStoreRetentionPeriodInHours: 168  # 7 days
        MagneticStoreRetentionPeriodInDays: 1825  # 5 year
      MagneticStoreWriteProperties:
        EnableMagneticStoreWrites: true # マグネティックストアへの書き込みを有効にする

aws cloudformation deploy \
    --template-file cfn.yaml \
    --stack-name Amazon-Timestream-Many-Data-Sample-use-Partition-Key-Stack

{
  "Dimensions": [
    {
      "Name": "deviceId",
      "Value": "device0001",
      "DimensionValueType": "VARCHAR"
    },
    {
      "Name": "sensorId",
      "Value": "sensor0001",
      "DimensionValueType": "VARCHAR"
    }
  ],
  "MeasureValueType": "MULTI",
  "MeasureName": "sensorData",
  "Time": "1745798400000",
  "TimeUnit": "MILLISECONDS",
  "MeasureValues": [
    {
      "Name": "temperature",
      "Value": "25.8",
      "Type": "DOUBLE"
    },
    {
      "Name": "humidity",
      "Value": "14.1",
      "Type": "DOUBLE"
    }
  ]
}

import boto3
import random
import json

from datetime import datetime, timedelta, timezone

from botocore.exceptions import ClientError

TIMESTREAM_DATABASE_NAME = "ManyDataSampleUsePartitionKeyDatabase"
TIMESTREAM_TABLE_NAME = "many-data-sample-use-partition-key-table"

client = boto3.client("timestream-write")

def main():
    JST = timezone(timedelta(hours=9), "JST")
    base_datatime = datetime(2025, 5, 23, 9, 0, 0, tzinfo=JST)
    base_unixtime = int(base_datatime.timestamp() * 1000)

    records = []

    # デバイス1〜3
    for device_number in range(1, 4):
        # センサー1〜6
        for sensor_number in range(1, 7):
            # 5年分のデータ（10分毎）
            for count in range(6 * 24 * 365 * 5):
                records.append(
                    make_record(
                        base_unixtime - count * 10 * 60 * 1000,  # 10分毎にする
                        f"device{device_number:04d}",
                        f"sensor{sensor_number:04d}",
                        round(random.uniform(0, 40), 1),  # 温度の乱数
                        round(random.uniform(0, 100), 1),  # 湿度の乱数
                    )
                )

    # Timestreamに書き込む（書き込み制限が最大100件のため、分割している）
    for i in range(0, len(records), 100):
        write_records(
            TIMESTREAM_DATABASE_NAME,
            TIMESTREAM_TABLE_NAME,
            records[i : i + 100],
        )

def make_record(unixtime, device_id, sensor_id, temperature, humidity):
    return {
        "Dimensions": [
            {"Name": "deviceId", "Value": device_id, "DimensionValueType": "VARCHAR"},
            {"Name": "sensorId", "Value": sensor_id, "DimensionValueType": "VARCHAR"},
        ],
        "MeasureValueType": "MULTI",  # マルチメジャーレコード
        "MeasureName": "sensorData",
        "Time": str(unixtime),
        "TimeUnit": "MILLISECONDS",
        "MeasureValues": [
            {
                "Name": "temperature",
                "Value": str(temperature),
                "Type": "DOUBLE",
            },
            {
                "Name": "humidity",
                "Value": str(humidity),
                "Type": "DOUBLE",
            },
        ],
    }

def write_records(database_name, table_name, records):
    try:
        resp = client.write_records(
            DatabaseName=database_name,
            TableName=table_name,
            Records=records,
        )
    except ClientError as e:
        if e.response["Error"]["Code"] == "RejectedRecordsException":
            print(e.response["Error"]["Message"])
            print(json.dumps(e.response["RejectedRecords"], indent=2))
        else:
            print(f"error: {e}")
        raise e

    # print(resp)

if __name__ == "__main__":
    main()

$ time python app.py 

python app.py  752.37s user 96.52s system 19% cpu 1:12:37.91 total

SELECT COUNT(*) AS total_records
FROM "ManyDataSampleUsePartitionKeyDatabase"."many-data-sample-use-partition-key-table"

SELECT COUNT(*) AS total_records
FROM "ManyDataSampleUsePartitionKeyDatabase"."many-data-sample-use-partition-key-table"
WHERE deviceId='device0001'

SELECT deviceId, sensorId,
	MAX_BY(time, time) AS latest_timestamp,
	MAX_BY(temperature, time) AS latest_temperature,
	MAX_BY(humidity, time) AS latest_humidity
FROM "ManyDataSampleUsePartitionKeyDatabase"."many-data-sample-use-partition-key-table"
WHERE deviceId='device0001'
GROUP BY deviceId, sensorId
ORDER BY latest_timestamp DESC, deviceId ASC, sensorId ASC

SELECT deviceId, sensorId,
	MAX(temperature) AS temperature_max,
	MIN(temperature) AS temperature_min,
	ROUND(AVG(temperature), 1) AS temperature_avg,
	MAX(humidity) AS humidity_max,
	MIN(humidity) AS humidity_min,
	ROUND(AVG(humidity), 1) AS humidity_avg
FROM "ManyDataSampleUsePartitionKeyDatabase"."many-data-sample-use-partition-key-table"
WHERE deviceId='device0001' AND time >= ago(30day)
GROUP BY deviceId, sensorId
ORDER BY deviceId ASC, sensorId ASC