DynamoDBで取得する列を指定した場合のキャパシティユニット - 闘うITエンジニアの覚え書き

DynamoDBで取得する列を指定して検索した場合のQueryのキャパシティユニットについて確認する。

AttributesToGet はレガシーパラメータだそうなので、以下のサンプルでは ProjectionExpression を使用する。
※ https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/dynamodb.html

import boto3
from boto3.dynamodb.conditions import Key
import json

dynamodb = boto3.resource('dynamodb')
table = dynamodb.Table('SampleTable')

def create_test_data():
    text_data = '1234567890' * 1000

    for i in range(1):
        for j in range(10):
            res = table.put_item(
                Item = { 
                    "pkey": f'pkey-{i+1:05d}',
                    "skey": f'skey-{j+1:05d}',
                    "column001": text_data,
                    "column002": text_data,
                    "column003": text_data,
                    "column004": text_data,
                    "column005": text_data,
                    "column006": text_data,
                    "column007": text_data,
                    "column008": text_data,
                    "column009": text_data,
                    "column010": text_data
                    },  
                ReturnConsumedCapacity = 'INDEXES'
            )   
            print(res)

def get_item(pkey, skey, columns):
    args = {}
    args['Key'] = {'pkey': pkey, 'skey': skey}
    args['ProjectionExpression'] = columns
    args['ReturnConsumedCapacity'] = 'INDEXES'
    res = table.get_item(**args)
    print(json.dumps(res['ConsumedCapacity'], indent=4))

def query(pkey, skey, columns):
    args = {}
    args['KeyConditionExpression'] = Key('pkey').eq(pkey)
    if skey:
        args['KeyConditionExpression'] &= Key('skey').eq(skey)
    args['ProjectionExpression'] = columns
    args['ReturnConsumedCapacity'] = 'INDEXES'
    res = table.query(**args)
    print(json.dumps(res['ConsumedCapacity'], indent=4))

def main():

    #create_test_data()

    print('### get_item(column * 5) ###')
    get_item('pkey-00001', 'skey-00001', 'column001, column002 ,column003, column004, column005')

    print('### get_item(column * 10) ###')
    get_item('pkey-00001', 'skey-00001', 'column001, column002 ,column003, column004, column005, column006, column007 ,column008, column009, column010')

    print('### query(column * 5) ###')
    query('pkey-00001', 'skey-00001', 'column001, column002 ,column003, column004, column005')

    print('### query(column * 10) ###')
    query('pkey-00001', 'skey-00001', 'column001, column002 ,column003, column004, column005, column006, column007 ,column008, column009, column010')

    print('### query(hash key only query - column * 5) ###')
    query('pkey-00001', None, 'column001, column002 ,column003, column004, column005')

    print('### query(hash key only query - column * 10) ###')
    query('pkey-00001', None, 'column001, column002 ,column003, column004, column005, column006, column007 ,column008, column009, column010')

if __name__ == '__main__':
    main()

結果は変わらない。レコードサイズ分のキャパシティユニットが消費される。

### get_item(column * 5) ###
{
    "TableName": "SampleTable",
    "CapacityUnits": 12.5,
    "Table": {
        "CapacityUnits": 12.5
    }
}
### get_item(column * 10) ###
{
    "TableName": "SampleTable",
    "CapacityUnits": 12.5,
    "Table": {
        "CapacityUnits": 12.5
    }
}
### query(column * 5) ###
{
    "TableName": "SampleTable",
    "CapacityUnits": 12.5,
    "Table": {
        "CapacityUnits": 12.5
    }
}
### query(column * 10) ###
{
    "TableName": "SampleTable",
    "CapacityUnits": 12.5,
    "Table": {
        "CapacityUnits": 12.5
    }
}
### query(hash key only query - column * 5) ###
{
    "TableName": "SampleTable",
    "CapacityUnits": 122.5,
    "Table": {
        "CapacityUnits": 122.5
    }
}
### query(hash key only query - column * 10) ###
{
    "TableName": "SampleTable",
    "CapacityUnits": 122.5,
    "Table": {
        "CapacityUnits": 122.5
    }
}

ドキュメントにも書いてある。
https://docs.aws.amazon.com/ja_jp/amazondynamodb/latest/developerguide/CapacityUnitCalculations.html#ItemSizeCalculations.Reads

あと、検索結果が０件の場合も、キャパシティユニットが消費される模様。

ただし、queryで複数件を一度に取得した時は１レコード毎に切り上げられず、
合計サイズでキャパシティユニットが計算されるので、少しお得。