DynamoDb is not a distributed hash map, it is a distributed forest of B-Trees. And B-Trees are what PG uses for indices and MySQL for both tables and indices. You don’t demoralize with it, but rather you build and maintain table and index like structures, they would contain submitted of ids or other data fields but it is the same with normal DB indexes. The difference if you have to maintain them manually. The upside is scale, serverlessnes and maybe less latency.