DynamoDB DAX Consistency SLA

Question

Hello AWS experts,

I understand that DynamoDB (and more specifically to my question, DAX) follows an "eventually consistent" data consistency model, and this means there is an expected delay between when some data is written and when a query of that same data will return the most recent information.

I've been reviewing the DAX and DynamoDB consistency models documentation (https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/DAX.consistency.html) and am trying to understand the expected time for data to reach consistency across all nodes in a DAX cluster. Specifically, I'm looking for an "order of magnitude" estimate. Is the expected time to consistency in the ballpark of 10ms, 100ms, 1000ms, or several seconds?  This is specifically related to the consistency of data which has been inserted/updated using the DAX cluster (no write-around or other ways of circumventing the DAX cache).

Any insights or references to further documentation would be greatly appreciated. Thank you!

Leeroy Hannigan · Accepted Answer

Hi,

The reason you cannot see this documented as DynamoDB or DAX do not provide an SLA for data consistency. In normal working conditions, both DynamoDB and DAX will be consistent within a 10-100ms. But as with any cluster based system, things can and do cause issue, so I would suggest that you do not take a dependency on consistency when using a caching solution such as DAX.

DynamoDB DAX Consistency SLA

Relevant content