Lets suppose I have a dynamodb table with 10 frequently accessed items of around 8KB each. I decided to use dax infront of the table. I got total 1 million read requests for the items.
a. Will I be charged for 10 dynamodb requests, since only 10 requests made it to dynamodb and rest were fetched from dax cache itself, or b. will I be charged for all 1 million dynamodb requests.
I had a similar question, and asked AWS. The answer I received was:
Whenever DAX has the item available (a cache hit), DAX returns the item to the application without accessing DynamoDB. In that case, the request will not consume read capacity units (RCUs) from DynamoDB table and hence there will not be any DynamoDB cost for that request. Therefore, if you have 10k requests and out of that if only 2k requests goes to DynamoDB, the total charge which gets charged will be for the 2k read request charge for DynamoDB, running cost for DAX cluster and data transfer charges (if applicable).
DynamoDB charges for DAX capacity by the hour and your DAX instances run with no long-term commitments. Pricing is per node-hour consumed and is dependent on the instance type you select. Each partial node-hour consumed is billed as a full hour. Pricing applies to all individual nodes in the DAX cluster. For example, if you have a three-node DAX cluster, you are billed for each of the separate nodes (three nodes in total) on an hourly basis.
