A social platform stores user posts in DynamoDB (managed NoSQL database, bills write capacity units per KB written — independently per table and per index). The table uses <code>userid</code>_ as the partition key and _<code>postid</code> (a ULID) as the sort key. The primary access patterns are: fetch all posts for a user (covered by the base table), and fetch posts by hashtag sorted by timestamp (not covered).
The team adds a Global Secondary Index (GSI — a separate index with its own partition key, billed independently from the base table for every write that touches its key attributes) with <code>hashtag</code> as the partition key and <code>created_at</code> as the sort key. Hashtag feed queries now work correctly.
Three months later, a DynamoDB cost review flags the table. Write costs have increased by 280% compared to before the GSI was added, even though write volume grew by only 15%.
What is causing the cost increase?
A) GSIs replicate item data to the index — large post items with media metadata cost more WCUs to replicate; but the billing mechanism is write units per KB, not a flat replication
surcharge, and item size affects WCU count the same way on both the base table and the GSI
B) The GSI has a hot partition on popular hashtags — write throttling on the GSI causes DynamoDB to retry writes internally; but DynamoDB does not bill for its own internal retry behavior — throttling causes errors and failures, not additional WCU charges
C) GSI writes are billed independently from the base table — every write that includes the GSI key attributes (<code>hashtag</code> + <code>created_at</code>) triggers a separate WCU charge for the GSI on top of the base table write, doubling or more the effective write cost per item
D) GSIs are cross-region by default — replication to other regions incurs data transfer charges billed as write units; but GSIs are always in the same region as the base table and have no cross-region behavior
Answer in the comments.
#AWS #SystemDesign #Database #DynamoDB #NoSQL

Smiley Face

Joud Awad

A DynamoDB cost puzzle: a social platform adds a Global Secondary Index (GSI) on hashtag and created_at to support hashtag feed queries. Three months later, write costs jump 280% despite only 15% write volume growth. The post presents four multiple-choice options explaining the cause, with the correct answer being that GSI writes are billed independently — every write touching the GSI key attributes incurs a separate WCU charge on top of the base table write, effectively doubling or more the write cost per item.

Day 17/30 AWS System Design Patterns

The answer is C.
DynamoDB (managed NoSQL database, bills write capacity units per KB per write destination) bills write capacity units for both the base table and each GSI (Global Secondary Index — billed independently per write) independently. When you write a post item that includes <code>hashtag</code> and <code>created_at</code> — the GSI key attributes — DynamoDB writes the item to the base table and also writes a projected copy to the GSI. Each write is billed separately.
The cost multiplier depends on item size. A 1 KB post item costs 1 WCU on the base table and 1 WCU on the GSI — doubling the write cost for that item. A 2.5 KB item costs 3 WCUs on the base table and 3 WCUs on the GSI. With one GSI, the effective write cost per item doubles for every item that includes GSI key attributes.
The 280% increase (not 200%) suggests most items are over 1 KB, or that item updates are modifying the GSI key attributes — which triggers a delete of the old GSI entry and a write of the new one, meaning two GSI write operations per update.
The fix: if only a fraction of posts have hashtags, don’t include <code>hashtag</code> on posts that don’t — DynamoDB only writes to the GSI when the key attributes are present in the item. If all posts have hashtags, reduce the GSI projection from ALL to only the attributes the query needs — smaller projected items cost fewer WCUs per GSI write.
<img src="https://media.daily.dev/image/upload/s--hFaIaBvS--/f_auto/v1786031505/ugc/content_2fcb0e77-b3f0-4ef6-9c58-0c9eb083e673?_a=BAMAMicg0" alt="ChatGPT Image Aug 6, 2026, 06_50_35 PM.png">

For a DynamoDB Indexes Deep Dive, please refer to my video: <a href="https://youtu.be/xVsEviV2vNA?si=ps2vhwQl1Ly-rEmn" target="_blank" rel="noopener nofollow">https://youtu.be/xVsEviV2vNA?si=ps2vhwQl1Ly-rEmn</a>

Also, it would mean a lot to me if you could support my content and stay in touch 🙏
<ul>
<li>YouTube: <a href="https://www.youtube.com/@system-design-lab" target="_blank" rel="noopener nofollow">https://www.youtube.com/@system-design-lab</a></li>
<li>LinkedIn: <a href="https://www.linkedin.com/in/joud-awad/" target="_blank" rel="noopener nofollow">https://www.linkedin.com/in/joud-awad/</a></li>
<li>Medium Blog: <a href="https://joudwawad.medium.com/" target="_blank" rel="noopener nofollow">https://joudwawad.medium.com/</a></li>
<li>Substack: <a href="https://joudawad.substack.com/" target="_blank" rel="noopener nofollow">https://joudawad.substack.com/</a></li>
</ul>

You are the_ real developer _brother.
Go ahead.