Think - with -Tech: Amazon DynamoDB

Monday, August 11, 2025

Amazon DynamoDB | Overview & Hands-On.

Amazon DynamoDB - Overview & Hands-On.

Scope:

Intro,
The Concept: DynamoDB,
Key terminology
Visual Srchitecture
Storage & Partitioning,
Capacity Modes,
Read & Write Internals,
Advanced Features,
Security,
Cost Drivers,
Common Use Cases,
Gotchas (things to watch out for) & Best Practices,
High-Level Architecture Flow,
Project: Hands-On.

Intro:

1. The Concept: DynamoDB

Amazon DynamoDB is AWS’s fully managed NoSQL key-value and document database designed for:

Low-latency (single-digit millisecond reads/writes)
Massive scalability (handles millions of requests/sec)
Serverless operation (no manual provisioning of servers)

NB:

DynamoDB is built on principles from Amazon’s internal Dynamo paper (2007), but with added features like on-demand scaling, global replication, and integrated security.

2. Visual Architecture

   + ---------------------- +
| Client Apps / API (Lambda, API GW) |
+----------+-----------+
v
              +----------------------+
| DAX Cluster (opt) |
| In-memory microsecond |
| latency cache |
+----------+-----------+
v
+---------+----------+
             | DynamoDB Table |
+---------+----------+
v
  | Partition Key Hash |
v
+---------------+---------------+
v   v v
+------+-----+ +------+-----+ +------+-----+
|Partition 1 | |Partition 2 | |Partition N |
|   SSD      | |   SSD | |   SSD      |
+------+-----+ +------+-----+ +------+-----+
v
| --> Streams --> Lambda / Kinesis   |
v
| Global Tables Replication   |
   v
+ ----------------+ ----------------+  --------------+
v v v
Region A Region B   .. Region N

3. Key terminology

Data Model

Table – top-level container for data.
Items – equivalent to a row in RDBMS.
Attributes – key-value pairs inside an item.
Primary Key – uniquely identifies each item:

Partition key only (simple primary key)
Example: UserID
Partition key + Sort key (composite primary key)
Example: UserID + OrderDate

Indexes

Primary Index – automatically created using the table’s primary key.
Local Secondary Index (LSI) – same partition key as the table, but different sort key.
Global Secondary Index (GSI) – different partition and sort key, can be queried independently.

4. Storage & Partitioning

Data is automatically sharded into partitions.
Partition key value → hashed to determine which physical partition stores it.
Partitions scale horizontally as throughput or storage needs grow.

5. Capacity Modes

Two modes determine cost and throughput:

Provisioned Capacity

twtech sets Read Capacity Units (RCUs) & Write Capacity Units (WCUs).
Optional Auto Scaling adjusts based on traffic.

On-Demand Capacity

Pay-per-request pricing.
Ideal for unpredictable workloads.

6. Read & Write Internals

Read Types

Eventually Consistent Reads – lower latency, cheaper (1 RCU = 2 reads/sec for 4KB items).
Strongly Consistent Reads – always return latest data (1 RCU = 1 read/sec for 4KB items).
Transactional Reads – part of ACID transaction.

Write Types

PutItem – insert or replace.
UpdateItem – modify existing.
DeleteItem – remove.
BatchWriteItem – write multiple items.

6. Advanced Features

Streams – real-time change logs for Lambda triggers or replication.
Global Tables – multi-region active-active replication.
TTL (Time-to-Live) – auto-delete expired items.
Transactions – ACID compliance for multiple items/tables.
Point-in-Time Recovery (PITR) – restore to any second in last 35 days.
DynamoDB Accelerator (DAX) – in-memory caching for microsecond reads.

7. Security

IAM Policies – control API access at table/item/attribute level.
KMS Encryption – at rest (default enabled).
VPC Endpoints – private network access.

8. Cost Drivers

RCUs/WCUs (provisioned mode) or request count (on-demand mode).
Storage (GB-month).
Data transfer across regions.
Optional features (DAX, Streams, PITR, Global Tables).

9. Common Use Cases

User profile stores
IoT telemetry ingestion
Gaming leaderboards
Session stores
Serverless backends
Real-time analytics with Streams + Lambda + Kinesis

10. High-Level Architecture Flow

11. Gotchas (things to watch out for) & Best Practices

Hot Partitions – avoid skewed partition key distribution.
Item Size Limit – 400 KB max.
Query vs Scan – use Query whenever possible (faster, cheaper).
Sparse Indexes – GSI with non-existent attributes only stores relevant items.
Write Sharding – append random suffix to hot keys to spread load.

Project: Hands-On

How twtech creates and use its DynamoDB tables for items.

Search in the aws services for: DynamoDB

Create tables with name: twtech-dynamodb-table

Create table: Table details

Capacity calculator: Average item size (KB)

Read/write capacity settings: Capacity mode

Warm throughput

NB:

Increasing the warm throughput value pre-warms the table to handle planned peak events without throttling or scaling delays.
By default, warm throughput values are visible for all tables and global secondary indexes.
These values automatically adjust as twtech increases its provisioned throughput or on-demand consumption without extra charges.
But if twtech chooses to change them manually, additional charges will apply.

Secondary indexes

Estimated read/write capacity cost

Here is the estimated total cost of provisioned read and write capacity for twtech table and indexes, based on its current settings.

Encryption at rest

All user data stored in Amazon DynamoDB is fully encrypted at rest.
By default, Amazon DynamoDB manages the encryption key, and twtech is not charged any fee for using it.

Deletion protection: turn on

Deletion protection is turned off by default.
Deletion protection protects the table from being deleted unintentionally.
twtech can turn on deletion protection during provisioning,or after the table has been created.