Performance in MongoDB and GridFS - mongodb

Question

I am developing a plugin that using mongodb. The plugin has to store some .dcm files (DICOM files) in the database as binary files. After that, the plugin has to store the metadata of the file and be able to make some query on only these metadata.
Naturally, I chose GridFs to answer at my problem. Because I can use the same file to store the binary data in the chunks collection and the metadata in the metadata field in the files collection (and bypass the sized limit of MongoDB).
But another problem comes to me. This solution would be great but I am storing at the same time the binary data and the metadata. Let me explain : first I store the binary file and after that I retrieve the file and read metadata from it and store the metadata in the same file. It is an obligation for me for some externals reasons. So I lost a lot of time to retrieve the file and restore it again. For update the metadata from a file that it is already stored, I am using this code :
GridFSDBFile file = saveFs.findOne(uri.getFileName());
if (file == null) {
return false;
} else {
file.setMetaData(new BasicDBObject());
file.save();
return true;
}
The main problem it that I have to find the file before to modify it and then store it AGAIN !
So my first question is : Is there a best way to retrieve file from the database instead of findOne(String fileName) ? Is the method findOne(ObjectID id) is faster ? (I don't think so because I think that fileName is already indexed by default, is not it ?)
I have tried another way to do it. To bypass this problem, I decided to store 2 different files, ones for binary data and ones for metadata. In this case, I don't loose time to retrieve the file in the database. But I have got 2 times more files... But I almost sure that it exist a better way to do it !
So my second question : Do you think that I would have to used 2 different collections ? One which used GridFs to store the binary data and the other one that used classic mongo storage (or GridFS) to only store the metada ?
Thank you a lot for reading me and for your answer :).

Answer 1

For your first question, both _id and filename fields are indexed by default. While _id field is unique, filename is not. So if you have files with same filenames, getting a file with the filename will be relatively slower than getting it by the _id field.
For your second question, you can always have metadata for any GirdFS file you inserted. That means you don't have to have more than GridFS. Use GridFS to insert data, but just before inserting it, assign your metadata to the file you want to insert. That way you can query files using the metadata. If the metadata you want to have is fixed for all documents, then you can have those fields indexed too, and queryable of course.