DynamoDB can get very expensive, very fast. The example below demostartes that and a quick learning point.

“If you were planning on storing YouTube views in Dynamo, you may calculate that each item is just a userId, videoId, and timestamp which is about 100 bytes. This means that you could store 10,000 views per second on a single shard. If you expect to have 10,000,000 views per second, you’d need ~1,000 shards to support the write load, which could cost approximately $150,000 per day based on current pricing”1

Understanding Capacity Units

Read Capacity Unit (RCU): 1 RCU = 1 strongly consistent read per second for items up to 4KB OR 2 eventually consistent reads per second for items up to 4KB

Write Capacity Unit (WCU): 1 WCU = 1 write per second for items up to 1KB

The Math Breakdown

Item Size Calculation: userId (string): ~20 bytes videoId (string): ~20 bytes timestamp: ~8 bytes DynamoDB overhead: ~50 bytes Total: ~100 bytes per view record

Write Capacity Needs: Each write consumes WCUs based on item size: ceiling(100 bytes / 1KB) = 1 WCU per view 10,000,000 views/second = 10,000,000 WCUs needed

Shard Requirements: Each shard maxes out at 1,000 WCUs 10,000,000 WCUs ÷ 1,000 WCUs per shard = 10,000 shards needed (The example says 1,000 shards, but that appears to be off by 10x)

Cost Calculation: 10,000,000 WCUs × $0.47 per million WCUs (approximate on-demand pricing) ~$4.70 per hour × 24 hours = ~$113 per day just for writes Add reads, storage, and other costs → easily $150k+ per day

Different item sizes:
The Rule:
1 WCU = 1 write per second for items up to 1KB
If your item is larger than 1KB, you need more WCUs

50 bytes   → ceiling(50/1024)   = ceiling(0.049) = 1 WCU
500 bytes  → ceiling(500/1024)  = ceiling(0.488) = 1 WCU  
1024 bytes → ceiling(1024/1024) = ceiling(1.0)   = 1 WCU
1025 bytes → ceiling(1025/1024) = ceiling(1.001) = 2 WCUs
2048 bytes → ceiling(2048/1024) = ceiling(2.0)   = 2 WCUs

Summary

DynamoDB is better suited for “medium-large” scale, not “internet-giant” scale. Batch multiple views into single items. Aggregate data before writing. Use different storage for high-volume, small-item scenarios.

The 1 WCU minimum per write, regardless of item size, is a key DynamoDB pricing characteristic that significantly impacts cost calculations for high-volume, small-item workloads. Even thought one view is 100 bytes, we end up using 1 WCU.

  1. System Design - DynamoDB ↩︎