which is faster(serialize/deserialize)using fast json or (select/insert)from/into database?

I see you already asked Mehdi Gholam a question in the comment section of his article on fastJSON here; why not ask him about performance/memory-use also.

Although, I think that performance/memory-use will vary depending on the various Types you are serializing, the database you use, etc.

As Mehdi states in his article, fastJSON will handle Dictionaries; you might ask him if Types that contain instances of the same Type have any problems being serialized ... since you are using a tree-structure.

You might also evaluate WCF (DataContract, DataMember) for serializing; I have serialized tree-structures with WCF. I first serialized with WCF to a memory stream; then used Gzip to compress it, and saved the compressed code to disk. The resulting Gzip file was almost always less than half the size compared to saving the WCF serialized code as XML.