After reading through over my other question, Using a Relational Database for Schema-Less Data, I started to question if your filesystem is much more appropriate than the usual relational database for storing and querying schemaless data.
As opposed to just creating a file system on the top of MySQL, why don't you just save the information straight to the filesystem? Indexing must be determined, but modern filesystems are extremely stable, have wonderful features like replication, snapshot and backup facilities, and therefore are flexible at storing schema-less data.
However, I can not find any good examples of somebody utilizing a filesystem rather than a database.
How can i find more assets regarding how to implement a schemaless (or "document-oriented") database like a layer on the top of the filesystem? Is anybody utilizing a modern filesystem like a schemaless database?
One factor you might want to give thought to Oracle's BFILE datatype, the industry pointer to some file on disk. Possibly that could be the very best of both mobile phone industry's? Microsoft SQL server does not appear to provide this capacity.
There is a large illustration of an implementation at Amazon's S3.
This kind of implementation is how many organisations are moving towards, since it scales essentially much better than a relational database can. The approach is straightforward, and delay pills work, as well as for some problems, it is a great solution. Within the situation of Amazon's S3, it is really nice for cloud storage, if you won't want to need to bother about the problems of storing the information yourself.
You are invited to have a look at our Solid File System, the industry virtual file system product with built-in support for file metadata and SQL-like search mechanism that searches through this data. Also please browse the article that describes the advantages of storing various kinds of data in different types of storages.
Yes a filesystem might be taken like a special situation of the NOSQL-like database system. It might possess some restrictions that needs to be considered throughout any design choices:
pros: - - simple, intuitive.
- uses many years of tuning and caching calculations
- easy backup, potentially easy clustering
items to consider:
richness of metadata - what kinds of data will it store, so how exactly does it let you query them, are you able to have hierarchal or multivalued characteristics
speed of querying metadata - not every fs's are particularly well enhanced with anything apart from size, dates.
lack of ability to become listed on queries (though that's virtually present with NoSQL)
inefficient storage usage (unless of course the file system works block suballocation, you'll typically blow 4-16K per item saved no matter size)
- Might not have the type of caching formula you would like for it's directory structure
- is commonly less tunable, etc.
- backup solutions might have trouble for the way you store things - too deep, a lot of products per node, etc - that might obviate an apparent benefit of this type of structure. securing for any LOCAL filesystem works pretty much obviously should you call the right programs, although not always for any network base fileesytem (individuals problems happen to be solved in a variety of ways, but it is certainly a design problem)