DynamoDB Interview Questions
Core Concepts
Q: What is DynamoDB? When to choose it?
DynamoDB is AWS's fully managed NoSQL key-value and document database. Serverless — no instances to manage, automatic scaling, single-digit millisecond performance at any scale.
Choose DynamoDB when:
•Need consistent single-digit ms latency at any scale
•Access patterns are known and simple (get by key, query by partition)
•Want serverless/fully managed (no capacity planning)
•Need automatic scaling for unpredictable traffic
•Multi-region active-active replication (Global Tables)
Don't use when:
•Complex queries with JOINs needed
•Flexible ad-hoc queries (use RDS/Aurora instead)
•Strong ACID transactions across many items (limited support)
•Analytical queries (use Redshift or Athena instead)
Q: DynamoDB data model — partition key, sort key, indexes.
python
import boto3
dynamodb = boto3.resource('dynamodb')
# Table with composite key: PK + SK
# PK = partition key (hash), SK = sort key (range)
table = dynamodb.create_table(
TableName='Orders',
KeySchema=[
{'AttributeName': 'customerId', 'KeyType': 'HASH'}, # Partition key
{'AttributeName': 'orderId', 'KeyType': 'RANGE'} # Sort key
],
AttributeDefinitions=[
{'AttributeName': 'customerId', 'AttributeType': 'S'},
{'AttributeName': 'orderId', 'AttributeType': 'S'},
{'AttributeName': 'orderDate', 'AttributeType': 'S'},
],
# GSI: query by different attributes
GlobalSecondaryIndexes=[{
'IndexName': 'OrdersByDate',
'KeySchema': [
{'AttributeName': 'customerId', 'KeyType': 'HASH'},
{'AttributeName': 'orderDate', 'KeyType': 'RANGE'}
],
'Projection': {'ProjectionType': 'ALL'}
}]
)
Key concepts:
•Partition key (PK): Determines which partition stores the item. High cardinality = even distribution.
•Sort key (SK): Within a partition, items sorted by SK. Enables range queries.
•GSI (Global Secondary Index): New PK/SK pair — query patterns not covered by table key.
•LSI (Local Secondary Index): Same PK, different SK — must be defined at table creation.
Q: DynamoDB single-table design.
python
# ENTITY TYPE OVERLOADING — store multiple entity types in one table
# PK = partition key, SK = sort key
# Pattern: PK="USER#123", SK="METADATA" → user info
# PK="USER#123", SK="ORDER#456" → user's order
# PK="ORDER#456", SK="ITEM#789" → order line item
# Benefits: single request fetches user + all orders (no JOIN)
# Get all items for user:
table.query(
KeyConditionExpression=Key('PK').eq('USER#123')
)
# Get user metadata only:
table.get_item(Key={'PK': 'USER#123', 'SK': 'METADATA'})
# Get user's orders by date range:
table.query(
KeyConditionExpression=Key('PK').eq('USER#123') &
Key('SK').between('ORDER#2024-01', 'ORDER#2024-12')
)
Q: DynamoDB capacity modes and auto-scaling.
Provisioned capacity: You specify read/write capacity units (RCU/WCU). Cheaper for predictable traffic. Auto-scaling adjusts within bounds.
•1 RCU = 1 strongly consistent read of ≤4KB/second
•1 WCU = 1 write of ≤1KB/second
On-demand capacity: Pay per request. No capacity planning. More expensive at steady scale. Best for unpredictable traffic.
DynamoDB Accelerator (DAX): In-memory cache, microsecond read latency. Sit in front of DynamoDB for read-heavy workloads.
Q: DynamoDB streams and event-driven patterns.
python
# DynamoDB Streams: capture every change (INSERT, MODIFY, REMOVE)
# → Trigger Lambda → fanout to other services
# Use cases:
# - Replicate to Elasticsearch for search
# - Audit log (who changed what, when)
# - Trigger notifications on certain changes
# - Cross-region replication (Global Tables uses streams internally)
# Global Tables: multi-region active-active
# Create table → enable Global Tables → add replica regions
# Writes replicate in < 1 second across regions
# Use for: disaster recovery, low-latency global reads/writes
Revision Notes
DYNAMODB: AWS managed NoSQL. Serverless. Single-digit ms at any scale.
Use for: known access patterns, key-value, auto-scaling, global tables
NOT for: complex queries, joins, ad-hoc analytics
DATA MODEL:
PK (partition key): determines storage partition, must be high cardinality
SK (sort key): range queries within partition
GSI: alternate PK/SK for different query patterns (can add after creation)
LSI: same PK, different SK (must define at table creation)
SINGLE-TABLE DESIGN: all entities in one table
PK="USER#123" SK="METADATA" | PK="USER#123" SK="ORDER#456"
Enables fetching related data in single request
CAPACITY:
Provisioned: specify RCU/WCU (cheaper, predictable) | On-demand: pay per request
1 RCU = 1 strong read ≤4KB | 1 WCU = 1 write ≤1KB
DAX: microsecond cache layer for reads
STREAMS: capture INSERT/MODIFY/REMOVE → trigger Lambda → event-driven
GLOBAL TABLES: multi-region active-active, <1s replication
