performance - How Fast is Windows File System for Database Operations? -
i came across crazy thought , wanted share , ask feasibility, performance wise:
the idea manage object database operations by:
- creating folder each class named after class name
- creating sub-folder each sub-class named after sub-class name
- creating file each object named after unique id
- creating sub-folder each index named after names of indexed fields
- creating shortcut file each index entry referring original object file
- reading/writing binary objects fast serializer/deserializer
- inserting/updating/deleting objects , index entries renaming object , shortcut files
- caching/paging using memory-mapped files
- querying utilize binary search on sorted file names
update: thank replies. thinking can improved using compression/encryption library such 7z, instead of dealing os file system. otherwise, of stated concerns far valid. i'm wondering kind of underlying file system does, example, oracle uses
cons:
- on filesystems, 1 byte file takes full block of 4kb. can huge problem depending of kind of object wanna store in database.
- most filesystems not designed scale directories containing millions of files.
- complex queries require opening/reading/deserializing/closing of million of files, , slow.
Comments
Post a Comment