DynamoDb partition key recommendation for Amazon Connect call meta data

Question

Scenario: A company is using Amazon Connect to receive high volume of calls. Assume there are about 3000 calls a day. For each call, they want to track certain data in a table in DynamoDb. Here are some example fields: Caller ID, Contact ID (globally unique id issued by Amazon Connect), time call started etc.

We are thinking of using the Contact ID as the partition key. This has high cardinality. However it is quite long. Here is a sample that I made up: 71a4aae6-1c6b-485e-8ecc-fd4782315a5c . We do not intent to query with this PK. We will have a separate Global Sec Index for querying.

Question: Is this a good idea? Are there any other recommendations specifically on how to design a table that is essentially holding telephony call logs with over 50 attributes and a fair bit of writes to those attributes when the call is underway.

Answer

The partitioning of the data would depend heavily on what you are trying to do with that data. Can you provide any more details on the use case? If the 50 attributes collected are used for reporting, no partitioning needed, I would argue you could get this data from an S3 bucket. What is the data used for? This would focus the attention of where you want to optimize your architecture using DDB design.

DynamoDb partition key recommendation for Amazon Connect call meta data

Relevant content