NeuralBlitz Amazon DynamoDB
What I do
I am a fully managed NoSQL database service by AWS, offering fast and predictable performance with seamless scalability. I support both key-value and document data models, providing single-digit millisecond latency at any scale. I offer built-in security, backup/restore, in-memory caching (DAX), and on-demand capacity modes. I am ideal for serverless architectures, high-traffic web applications, and distributed systems requiring elastic scaling.
When to use me
- Serverless applications using AWS Lambda
- High-traffic web applications and APIs
- Gaming leaderboards and real-time analytics
- Session storage and user state management
- IoT data ingestion and time-series storage
- E-commerce product catalogs
- Mobile backends with offline sync
- Microservices requiring independent data stores
- Applications needing global tables for multi-region replication
Core Concepts
- Tables, Items, and Attributes: Data organized in tables containing items with multiple attributes
- Primary Key: Either simple (partition key) or composite (partition + sort key)
- Partition Key and Sort Key: Determine data distribution and sorting within partitions
- Local Secondary Indexes (LSI): Alternative sort key on same partition key
- Global Secondary Indexes (GSI): Alternative partition/sort key combinations for different access patterns
- DynamoDB Streams: Capture item-level changes for event-driven architectures
- DAX (DynamoDB Accelerator): In-memory cache providing microsecond read latency
- On-Demand vs Provisioned Capacity: Flexible billing modes for varying workloads
- Time to Live (TTL): Automatic item expiration for managing data lifecycle
- Global Tables: Multi-region, multi-active replication for global low-latency access
Code Examples
Basic Operations
import boto3
from boto3.dynamodb.conditions import Key, Attr
from botocore.exceptions import ClientError
dynamodb = boto3.resource("dynamodb", region_name="us-east-1")
table = dynamodb.Table("Products")
def create_product(product_data):
table.put_item(Item=product_data)
return product_data["product_id"]
def get_product(product_id):
response = table.get_item(Key={"product_id": product_id})
return response.get("Item")
def update_product(product_id, updates):
update_expr = "SET " + ", ".join([f"{k} = :{k}" for k in updates.keys()])
expr_attr_vals = {f":{k}": v for k, v in updates.items()}
response = table.update_item(
Key={"product_id": product_id},
UpdateExpression=update_expr,
ExpressionAttributeValues=expr_attr_vals,
ReturnValues="UPDATED_NEW"
)
return response["Attributes"]
def delete_product(product_id):
response = table.delete_item(
Key={"product_id": product_id},
ReturnValues="ALL_OLD"
)
return response.get("Attributes")
def batch_write_products(products):
with table.batch_writer() as batch:
for product in products:
batch.put_item(Item=product)
def batch_get_products(product_ids):
keys = [{"product_id": pid} for pid in product_ids]
response = table.batch_get_item(Keys=keys)
return response.get("Responses", {}).get("Products", [])
Query and Scan Operations
def query_products_by_category(category, limit=50):
response = table.query(
KeyConditionExpression=Key("category").eq(category),
Limit=limit
)
return response.get("Items", [])
def query_products_with_filter(category, min_price=None, in_stock=None):
key_cond = Key("category").eq(category)
filter_expr = None
if min_price:
filter_expr = Attr("price").gte(min_price) if not filter_expr else filter_expr & Attr("price").gte(min_price)
if in_stock is not None:
filter_expr = Attr("in_stock").eq(in_stock) if not filter_expr else filter_expr & Attr("in_stock").eq(in_stock)
response = table.query(
KeyConditionExpression=key_cond,
FilterExpression=filter_expr,
Limit=100
)
return response.get("Items", [])
def scan_all_products():
items = []
response = table.scan()
items.extend(response.get("Items", []))
while "LastEvaluatedKey" in response:
response = table.scan(ExclusiveStartKey=response["LastEvaluatedKey"])
items.extend(response.get("Items", []))
return items
def scan_with_filter(min_price=100, max_price=500):
response = table.scan(
FilterExpression=Attr("price").between(min_price, max_price) & Attr("in_stock").eq(True),
ProjectionExpression="product_id, name, price, category"
)
return response.get("Items", [])
def query_with_pagination(category, start_key=None, page_size=20):
kwargs = {
"KeyConditionExpression": Key("category").eq(category),
"Limit": page_size
}
if start_key:
kwargs["ExclusiveStartKey"] = start_key
response = table.query(**kwargs)
return response.get("Items", []), response.get("LastEvaluatedKey")
def query_with_sort_range(category, sort_key, operator="gte"):
operators = {"gte": ">=", "lte": "<=", "gt": ">", "lt": "<", "eq": "="}
key_cond = Key("category").eq(category) & Key("created_at")[operator](sort_key)
response = table.query(KeyConditionExpression=key_cond)
return response.get("Items", [])
Secondary Indexes
def create_indexes():
table = dynamodb.Table("Products")
table.update(
AttributeDefinitions=[
{"AttributeName": "category", "AttributeType": "S"},
{"AttributeName": "price", "AttributeType": "N"},
{"AttributeName": "SKU", "AttributeType": "S"}
],
GlobalSecondaryIndexUpdates=[
{
"Create": {
"IndexName": "CategoryPriceIndex",
"KeySchema": [
{"AttributeName": "category", "KeyType": "HASH"},
{"AttributeName": "price", "KeyType": "RANGE"}
],
"Projection": {"ProjectionType": "ALL"},
"ProvisionedThroughput": {"ReadCapacityUnits": 5, "WriteCapacityUnits": 5}
}
}
]
)
def query_by_SKU(SKU):
table = dynamodb.Table("Products")
response = table.query(
IndexName="SKUIndex",
KeyConditionExpression=Key("SKU").eq(SKU)
)
return response.get("Items", [])[0] if response.get("Items") else None
def query_products_by_gsi(category, limit=100):
table = dynamodb.Table("Products")
response = table.query(
IndexName="CategoryPriceIndex",
KeyConditionExpression=Key("category").eq(category),
ScanIndexForward=True,
Limit=limit
)
return response.get("Items", [])
Conditional Updates and Transactions
def update_with_condition(product_id, new_price, expected_version):
try:
response = table.update_item(
Key={"product_id": product_id},
UpdateExpression="SET price = :price, version = :new_version",
ConditionExpression="version = :expected_version",
ExpressionAttributeValues={
":price": new_price,
":expected_version": expected_version,
":new_version": expected_version + 1
},
ReturnValues="UPDATED_NEW"
)
return True, response["Attributes"]
except ClientError as e:
if e.response["Error"]["Code"] == "ConditionalCheckFailedException":
return False, None
raise
def update_inventory(product_id, quantity_change):
return table.update_item(
Key={"product_id": product_id},
UpdateExpression="SET inventory_count = if_not_exists(inventory_count, :zero) + :change",
ExpressionAttributeValues={":zero": 0, ":change": quantity_change},
ConditionExpression="attribute_exists(product_id)",
ReturnValues="UPDATED_NEW"
)
def transfer_points(from_user, to_user, amount):
dynamodb_transact = boto3.resource("dynamodb").Table("UserPoints")
try:
response = dynamodb_transact.transact_write_items(
TransactItems=[
{
"Update": {
"TableName": "UserPoints",
"Key": {"user_id": from_user},
UpdateExpression="SET points = points - :amount",
ConditionExpression="points >= :amount",
ExpressionAttributeValues={":amount": amount}
}
},
{
"Update": {
"TableName": "UserPoints",
"Key": {"user_id": to_user},
UpdateExpression="SET points = points + :amount",
ExpressionAttributeValues={":amount": amount}
}
}
]
)
return True
except ClientError as e:
return False
def atomic_counter(table_name, counter_key, amount):
table = dynamodb.Table(table_name)
response = table.update_item(
Key={"id": counter_key},
UpdateExpression="SET #cnt = #cnt + :amount",
ExpressionAttributeNames={"#cnt": "count"},
ExpressionAttributeValues={":amount": amount},
ReturnValues="UPDATED_NEW"
)
return response["Attributes"]["count"]
Streams and TTL
def enable_ttl(table_name, ttl_attribute="expires_at"):
dynamodb_client = boto3.client("dynamodb")
dynamodb_client.update_time_to_live(
TableName=table_name,
TimeToLiveSpecification={
"Enabled": True,
"AttributeName": ttl_attribute
}
)
def set_item_with_ttl(item, ttl_seconds=86400):
from datetime import datetime, timedelta
import time
item["expires_at"] = int(time.time() + ttl_seconds)
table.put_item(Item=item)
def create_ddb_stream_handler(lambda_function_name):
import boto3
dynamodb_client = boto3.client("dynamodb")
streams_client = boto3.client("dynamodbstreams")
response = dynamodb_client.describe_table(TableName="Products")
stream_arn = response["Table"]["LatestStreamArn"]
lambda_client = boto3.client("lambda")
lambda_client.create_event_source_mapping(
EventSourceArn=stream_arn,
FunctionName=lambda_function_name,
StartingPosition="LATEST",
BatchSize=100,
MaximumBatchingWindowInSeconds=60
)
def process_stream_records(records):
for record in records:
event_name = record["eventName"]
item = record["dynamodb"].get("NewImage")
old_item = record["dynamodb"].get("OldImage")
if event_name == "INSERT":
print(f"New item added: {item}")
elif event_name == "MODIFY":
print(f"Item modified: {old_item} -> {item}")
elif event_name == "REMOVE":
print(f"Item removed: {old_item}")
def backup_table(table_name, backup_name):
dynamodb_client = boto3.client("dynamodb")
dynamodb_client.create_backup(
TableName=table_name,
BackupName=backup_name
)
def restore_from_backup(backup_arn, new_table_name):
dynamodb_client = boto3.client("dynamodb")
dynamodb_client.restore_table_from_backup(
TargetTableName=new_table_name,
BackupArn=backup_arn
)
Best Practices
- Choose Primary Key Wisely: Avoid hot partitions by distributing data across partition keys
- Use GSIs for Multiple Access Patterns: Create GSIs for queries that don't fit primary key access
- Implement Proper Error Handling: Use exponential backoff and jitter for retry logic
- Use Batch Operations: BatchGetItem and BatchWriteItem for bulk operations reduce API calls
- Choose Capacity Mode Appropriately: Use on-demand for variable workloads; provisioned for predictable patterns
- Enable DAX for Read-Heavy Workloads: DAX provides sub-millisecond read latency for hot data
- Implement TTL for Data Expiration: Use TTL instead of scheduled deletes for automatic cleanup
- Use Projections to Reduce Data Transfer: Specify only needed attributes in queries
- Leverage Adaptive Capacity: DynamoDB automatically handles hot partitions; design accordingly
- Monitor with CloudWatch: Track throttled requests, consumed capacity, and latency metrics