How to store json data in jquery?

How NoSQL Databases like MongoDB or reThinkDB store and retrieve JSON data from disk?

• I am planning to learn a new language and I am being provoked to implement a simplest NoSQL Db Engine in it. I have personally used MongoDB and loved it. As people know it's based on JSON(BSON), I am interested to know how the internals work. How data is retrieved from JSON based Objects. What would be the algorithm and how performance and efficiency was taken place?

• Answer:

  So the internals of MongoDB are not that complicated. If you look around, you should be able to find some talks on-line that cover the exact details. Of course, you can always read the source code too :) That said, here's a quick overview. MongoDB uses memory-mapped files. This is an Operating System construct that maps disk locations to spots in memory. Effectively, MongoDB simply pretends that the whole DB is in memory and lets the OS swap files to/from disk. MongoDB uses the "fsync" command to tell the OS when it is ready to flush to disk. This has some efficiency and reliability concerns so their is also a journal file that keeps track of actions performed. When more space is required, MongoDB simply creates a new file (up to 2gb) and maps it into memory. Each Database has a set of files named after if (mydb.1, mydb.2, mydb.3, etc). The different collections within a DB all share the same files. Documents are written to disk using the http://bsonspec.org/ format. To be clear, they are written to memory locations and flushed to disk. To allow for changes in the size of the documents, they are often padded. If you look at the collection statistics you can see the "padding factor". Documents also form a doubly-linked list. They have pointers to "next" & "prev" to allow for walking of a collection. Indexes are written to the file in blocks using B-trees. They have many of the same basic details and limitations as those indexes used in relational databases like MySQL. MongoDB uses a custom wire format. So commands arrive in a very specific format that is not really BSON. The details for this can be found in the driver docs. If you want to get into replication and sharding those are whole other complicated issues. And frankly, MongoDB does those quite poorly. All in all, MongoDB "works", but is not particularly awesome. It has lots of poor architecture decisions that still plague it today and that make it hard to operate large clusters. Some ideas to consider, things you could improve: Memory-mapped files are "easy", but they're not particularly safe: http://ayende.com/blog/162791/on-memory-mapped-files The BSON format is really inefficient. It takes up a lot of space as it stores both keys and values with no compression. Loading up a document requires loading the entire document into memory to access any part of the document. There are lots of possible variants on B-trees that are used to optimize performance, especially for things like "count" operations. Those are things you need to design in advance. There is no compression of data. This sounds like a great spot for improvement. MongoDB uses a single write-lock per DB. You should consider how much concurrency you want to allow.