DAX does not have Query cache hits

Question

Hello,

I have an architecture with 2 Lambdas, a DynamoDB table with PK+SK, and DAX. The use-case is as follow:
- Lambda 1 put Items in DAX using PK and SK with standard DynamoDB interface

```
        Table table = dynamoClient.getTable(tableDetails.getTableName());
        PrimaryKey itemKey = new PrimaryKey();
        itemKey.addComponent(tableDetails.getTablePKName(), tableDetails.getTablePKValue());
        itemKey.addComponent(tableDetails.getTableSKName(), tableDetails.getTableSKValue());
        Item item = new Item().withPrimaryKey(itemKey).withJSON("document", jsonResult);
        PutItemOutcome output = table.putItem(item);
```

- Lambda 2 query same items from DAX using only PK with standard DynamoDB interface

```
       Table table = dynamoClient.getTable(tableDetails.getTableName());
       QuerySpec spec = new QuerySpec().withHashKey(tableDetails.getTablePKName(), tableDetails.getTablePKValue());
       ItemCollection items = table.query(spec);
```
None of the inserted items was detected in DAX by the query. The Metrics are for 10 instances of Lambda 1 and Lambda 2: 
- PutItem requests : 10
- Query requests : 10
- Query cache misses : 10
- Query cache hits : 0

Why DAX does not have Query cache hits in this use-case ? The use-case does not support query with PK+SK. Are there some conditions for java code  in this use-case ?

Update: Lambda 2 is a destination for Lambda 1, so for a PK, Lambda 2 runs always after Lambda 1.

Thank you,  
Mihai ADAM

Accepted Answer

A simple answer to your question is that you are expecting write operations (Lambda 1) to update the Query cache, but it will not. Query cache is only populated from `Query` and `Scan` operations.

- When you write to DAX, the items are cached in the `Item` cache, using a key (TableName, PK, SK?).  
- When you do a Query from DAX, you read from the `Query` cache, which uses the requests parameters in the request (TableName, KeyConditionExpression, ProjectionExpression, ExclusiveStartKey, Limit, among others).

From that information you should be able to understand why you are not getting cache its.

Answer

Here are some points to consider that might explain why you're not seeing query cache hits in your use case:
* DAX caches GetItem, BatchGetItem, Query, and Scan results, but exact match queries are required for cache hits.
* DAX uses a write-through strategy, directly caching PutItem operations but not automatically caching Query results unless the exact query has been made before.
* Querying only by PK generates a different cache key than inserting with PK+SK, leading to cache misses.
* Even if Lambda 2 runs after Lambda 1, the absence of SK in the query changes the cache key, causing cache misses.
* DAX configuration and the TTL of cached items can affect cache hit rates. Frequent updates or deletes can invalidate cache items.

So, to improve cache hits, you need to align query patterns with caching behavior, consider a cache warming strategy, and adjust DAX configurations for better cache utilization.

DAX does not have Query cache hits

Relevant content