Querying across a time range

We have the need to query events stored in our table across a time range. A user could ask for events from a specfic time in one day, to another specfic time in another day.

We have been thinking about creating our partition key as a truncated value, possibly the day. That way with one query we could get a full day of events and then do some client side filtering for the specfic time range using the full time attribute.

Now this may be a problem if we have to query across multiple days, meaning multiple queries. A day could habe anything from a few thousand events, to a couple of hundred thousand.

Is the truncated timestamp (down to a day) the best approach, or is there a better one?

We were also wondering could there be a better aws serverless solution for us to use for this use case, while still using dynamodb for event storage.

The best way to achieve it is by having a static value as the gsi Partition key and a timestamp as the sort key.

Now you can efficiently query for any data range, be it a specific minute of a day without filtering.

Be mindful that keeping a single value as a partition key of your index limits your writes to approx 1000 WCU per second. If you intend to scale beyond that, consider sharding that GSI value.