Design and Implementation of The Metadata Modification Concept Minimizing File Modification

摘要

In recent years, the capacity of files such as documents and multimedia has been greatly increased. The user not only stores and reads data, but also performs various activities such as adding or deleting contents to an existing file. For editing, the file system copies all the contents of the disk into memory, and the user modifies this data using the application. Then, when storing the data again, the user application is using the write function to rewrite it all over the disk from beginning to end. In the vim editor, even when some data is modified, all data is saved again, resulting in slower performance as the file size increases. An example of another method is tar. tar consists of several sub-files, and if you delete one of these sub-files, all the sub-files after the deleted file are pulled forward and rewritten. This approach requires a lot of disk-write requests despite the small number of data changes, and it takes a long time to execute. In this paper, we proposed the Metadata Modification technique, which is a new concept of modifying part of a file quickly through some metadata modification. In this modification we will use a function call sys_meta_modification(). This function is a prototype, assuming a situation in which a portion of the file is deleted, and although it is a limited function, it treats the file in a way that is fundamentally different from the API of the original primitive file system. This function modifies the pointer in the metadata when the data is deleted on a block-by-block basis and returns the block to be deleted to the kernel. In other words, because there is no need to write back data to a disc from beginning to end, it was very quick to delete the contents of the file. It also shows that if you create a user application that imitates tar using this system call and deletes the sub-file, the data will be deleted quickly.