Common Causes of Write Throttling in DynamoDB
Question
An application stores its data in a DynamoDB table.
The application team finds that they have a new query pattern, creating a new LSI to accommodate the new queries.
The application starts to experience throttling when performing write operations on the base table even though the number of table updates has not increased.
What is the cause of the issue?
Answers
Explanations
Click on the arrows to vote for the correct answer
A. B. C. D.Answer: B.
Option A is incorrect because read capacity units are not related to write throughput.
Option B is CORRECT because LSI and base table share capacity units.
Performing write operations to a table causes the LSI to be also updated.
This consumes write capacity units from the base table.
In this case, the addition of the LSI is causing the provisioned write capacity to be exceeded.
Option C is incorrect because latency would not be causing request throttling.
Option D is incorrect because consistency level is associated with read operations, not write operations.
Reference:
https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/SecondaryIndexes.htmlThe most likely cause of the throttling issue experienced by the application during write operations on the DynamoDB table, after creating a new LSI to accommodate a new query pattern, is option B - "LSI shares write-capacity with the base table."
When a new LSI is created, it shares the same underlying storage resources as the base table, including read and write capacity. While the LSI is optimized for a specific query pattern, any write operations on the base table or the LSI can affect the overall write capacity of the table.
Therefore, creating a new LSI can cause write capacity to be spread thin between the base table and the new LSI, leading to throttling of write operations on the base table. This can happen even if the number of updates to the base table has not increased because the LSI also consumes write capacity.
Option A, "LSI shares read-capacity with the base table," is incorrect because LSIs have their own read capacity, and do not share it with the base table.
Option C, "Data replication to the LSI is synchronous causing latency," is incorrect because data replication to an LSI is asynchronous, meaning there should be minimal latency incurred.
Option D, "Creation of the LSI forces write operations to be strongly consistent," is also incorrect because creating an LSI does not have any impact on the consistency model used for write operations.
