DynamoDB vs Elasticache vs Google guava

We store some data in DynamoDB. An entry from the table will be queried during every request to our service. Each entry will be of size 1000 bytes, and the size of entire table will be of 1GB. We are looking to reduce DynamoDB cost by using cache, and possibly reduce the latency as well.

We seem to have 3 solutions depending upon the performance benchmarks and cost of DynamoDB, Elasticache and Google Guava cache.

• Use DynamoDB only


• Use a distributed cache (Elasticache) and fall back to DynamoDB on a cache miss.


• Use a local cache and fall back to DynamoDB on a cache miss. This is the lease preferred option as local cache supports TTL eviction only. We would prefer to invalidate the cache as soon as the data is updated in DynamoDB.

To make a decision, I need performance benchmarks of DynamoDB, Elasticache and Google guava cache. I am surprised I couldn't find it in Google search. Can someone give some idea?