Hi,
Is there any way for concatenating (particularly large video files) files without having to copying one after the other. But, manipulating the file entries in the directory and joining the second file's cluster chain to the first file's chain.

This facility allows us to fetch a large (movie) file from any offset and write into a chunk file and when all chunks are complete we can quickly join them together into one file.Or even when two consecutive chunks are fully fetched we can join them into one chunk.

Thanks

Isn't this pretty much how torrents work? They pre-allocate filespace on your filesystem, and then as each chunk is downloaded, it is inserted into the empty file at the correct offset until the file is complete.

So, maybe take a look at the sourcecode of any torrent up/downloading program, and you'll see how it's done.

You probably won't want to do directory-entry manipulations directly to the disk, because different disks may be using different filesystems, which will have very different methods for handling directory structures and it'll all get very complicated very quickly.

The idea was not to pre-allocate space as movie files are around 6-8 GB size. Moreover user may or may not completely download and watch it.He may just see portions of it (streaming VOD contents & and ability to play from any chapter like in DVD, but without fetching complete file). Therefore, we would like to create chunk files during random access downloads and join them later.

Thanks

Linux supports many types of filesystems. Even if you could do this at the filesystem level for one, you won't be able for another.
Unix systems are also inode based. That means you might have deleted files which still occupy disk space as long as a process has it open. This is something to keep in mind in writing your app. Look at `lsof +L1' for examples.

One thing to look into is using sparse files.

Indeed Sparse Files meets my requirement. It does not allocate data blocks and waste disk space.