So far all the articles I read about Cassandra all mentions about doing data denormalizing/duplication to improve read performance (e.g. [ebay blog](http://www.ebaytechblog.com/2012/07/16/cassandra-data-modeling-best-practices-part-1/ and cassandra blog). But to me it seems like the use case is only if you are using Cassandra as the main database for querying.
Currently I have ElasticSearch indexed my Cassandra DB (where everything is still normalized), so does it make sense for me to still denormalize my Cassandra DB given that all my queries actually go through ElasticSearch (ie. it will return list of ids and I fetch the ids directly from Cassandra)?