I am considering to serialize a big set of database records for cache in Redis, using python and Cassandra. I have either to serialize each record and persist a string in redis or to create a dictionary for each record and persist in redis as a list of dictionaries.
Which way is faster? pickle each record? or create a dictionary for each record?
And second : Is there any method to fetch from database as list of dic's? (instead of a list of model obj's)
Instead of serializing your dictionaries into strings and storing them in a Redis LIST (which is what it sounds like you are proposing), you can store each dict as a Redis HASH. This should work well if your dicts are relatively simple key/value pairs. After creating each HASH you could add the key for the HASH to a LIST, which would provide you with an index of keys for the hashes. The benefits of this approach could be avoiding or lessening the amount of serialization needed, and may make it easier to use the data set in other applications and from other languages.
There are of course many other approaches you can take and that will depend on lots of factors related to what kind of data you are dealing with and how you plan to use it.
If you do go with serialization you might want to at least consider a more language agnostic serialization format, like JSON, BSON, YAML, or one of the many others.