algorithm - How to model many to many relation in Cassandra 3.0 with performance in mind -

i trying define cassandra tables following problem.
have following relations:
user (user_id) can see many adds (add_id).
add (add_id) can saw more 1 user.

and want make following querys in cassandra:

1. given user (user_id) give me adds (add_id) user saw
2. given add (add_id) give me users (user_id) saw add

the aproach took create 2 tables whith relations

create table adds_by_user (     user_id text,     add_id text,     primary key (user_id, add_id) );  create table user_by_add (     add_id text,     user_id text,     primary key (add_id, user_id) ); 

my algorithm following
given user_id:

1. give me add_id user watched.
2. for each add_id, give me user_id watched add_id
3. for each user_id, give me add_id each user watched

another graphical way see is:

(one) user_id -> (many) add_id -> (many+) user_id -> (many++) add_id

the problems arises when have lot of relations , algorithm spends lot of time doing queries (despite fact each individual query done in 0.5 milliseconds or less, but there lot of them)

i using async queries

is there model can make queries faster?

is there model make algorithm make less queries , respect cassandra standards?

thanks!!

example of data model:

create table user_address_user(     user_id text,     address_id text,     user_level2 text,     address_level2 text,      primary key((user_id), address_id, user_level2, address_level2) ); 

this table structure can seen as

map<user_id, sortedmap<address_id, sortedmap<user_level2, sortedset<address_level2>>>>