Question

I have a set of documents that are the same but could be sorted into two distinct groups based on usage.

One of these groups, we’ll call them “current”, has a low volume of documents being queried a lot.

The other group, we’ll call them “backlog”, is a huge volume of documents queried much less often.

My thought was that not mixing these these two types of the same document would allow me to query the very active “current” docs without needing to sift through the huge amount of “backlog” every time.

Should these be in two different collections or the same “cars” collection?

mongo.collection('cars', function(err, cars){
    cars.find({type:'new', color:'blue'}).toArray(function(err, newBlueCars) {
        //do something with newBlueCars
    });
});

mongo.collection('cars', function(err, cars){
    cars.find({type:'used', color:'blue'}).toArray(function(err, usedBlueCars) {
        //do something with usedBlueCars
    });
});

OR

mongo.collection('cars.current', function(err, cars){
    cars.find({color:'blue'}).toArray(function(err, currentBlueCars) {
        //do something with newBlueCars
    });
});

mongo.collection('cars.backlog', function(err, cars){
    cars.find({color:'blue'}).toArray(function(err, backlogBlueCars) {
        //do something with usedBlueCars
    });
});

Answer 1

I figured I would post an actual answer, now that it has been sufficiently hashed out in comments (the other answer isn’t quite good enough).

The benefits of keeping it all in one collection are:

• Can query all records in a single query
• Can use sparse index to efficiently index the “current” records

The downsides of keeping it all in one collection are:

• Indexes probably need to have the “current” field leading all of them, which takes extra space and makes lookups slower
• Can’t index multiple fields with a sparse index

The benefits of splitting them into two collections are:

• You get a free index, so you don’t need to lead all of your indexes with the “current” field
• Your “current” data will have much smaller indexes that can easily fit in RAM

The downsides of splitting them into two collections are:

• You’ll need to run two queries to do anything with the dataset as a whole, and merge the results