Redis vs. Cassandra for high volume database query

[donboscow]

I have a scenario - a shop runs hundreds and thousands of queries every hour. Each transaction has values like merchant id client id, purchase id, item name, item type, price, time of purchase, etc.

Potentially, the database can have millions of such records, and that can pretty much scale upto billions as well. The web front end will support queries like "find the records of transactions by that customer id brought on this date within the price range this to that", etc. The response must be near real time, say not slower than what time Google typically takes to return a search result.

Keeping in mind cost, and most importantly, performance, which is preferable - Redis, or a NoSQL DB like MondoDB or Cassandra? Will Redis offer significantly more performance benefits because it is in-memory, or will the difference be negligible in this situation?