The Cloud Economist
Posts
How Samsung Cloud Uses DynamoDB To Optimize Costs

How Samsung Cloud Uses DynamoDB To Optimize Costs

Some valuable strategies that are easy to implement and that you can use yourself to save on database costs.

Uriel Bitton
December 15, 2024

Samsung Cloud & DynamoDB to optimize database costs

When Samsung Cloud transitioned its database from Apache Cassandra to DynamoDB, they didn’t just gain scalability — they also significantly optimized costs.

By adopting a few smart strategies, Samsung achieved a 40% reduction in total cost of ownership while scaling to 3.5 PB of data and handling over 100 billion daily reads and writes.

If you’re looking to reduce your DynamoDB costs without sacrificing performance, this post will break down 5 cost-saving strategies Samsung used — strategies you can apply to your own workloads.

1. Solid Data Modeling: Switch to Single Table Design

When Samsung first migrated to DynamoDB, they used one entity per table. While this worked initially, they later adopted the Single Table Design, which brought significant benefits:

• Lower Throughput Costs: Related data is queried together, reducing latency and request overhead.

• Reduced Operational Costs: Sharing provisioned capacity across workloads minimized throttling risks, cutting down on table management overhead.

2. Auto Scaling Policies

Scaling tables manually is tedious. To simplify operations, Samsung implemented unified auto-scaling policies. By leveraging DynamoDB’s adaptive capacity and burst features, they efficiently handled uneven access patterns and sudden traffic spikes without over-provisioning resources.

This approach eliminated manual intervention while ensuring cost-efficient scaling.

3. Time to Live (TTL) for Expired Data

Before TTL was introduced, Samsung relied on batch jobs to delete expired data — an expensive and time-consuming process.

With TTL, Samsung achieved dramatic results:

• 94% Storage Reduction: Tables shrank from 1.2 PB to 74.5 TB.

• Reduced RCU and WCU Costs: RCUs dropped by 60%, WCUs by 70%.

• 90% Cost Reduction for TTL-enabled tables: Expired data was deleted for free, saving millions of dollars annually.

4. Reserved Capacity for Predictable Workloads

Samsung purchased reserved capacity to lock in lower prices for predictable workloads. By accurately forecasting usage, they achieved up to 50% annual discounts.

Even when they overestimated usage, the cost benefits outweighed the risks. If your application has consistent traffic patterns, reserved capacity is an easy way to slash costs.

5. Standard-Infrequently Accessed (IA) Tables

In 2021, Samsung began migrating less-used tables to the Standard-IA table class, which reduces storage costs by up to 60%.

By analyzing table costs and usage patterns, they identified 50 candidate tables for migration. This switch required zero code changes and didn’t compromise performance, delivering a 30% cost reduction for those tables.

The Takeaway

Samsung Cloud’s DynamoDB cost optimization journey highlights five actionable strategies:

1. Efficient data modeling with Single Table Design

2. Automating scaling with unified policies

3. Using TTL to eliminate expired data for free

4. Purchasing reserved capacity for predictable savings

5. Leveraging the Standard-IA table class for low-access tables

By adopting these practices, Samsung achieved massive cost savings while scaling their workloads globally.