File System Design part 1: XFS
Finally, lets look at the XFS superblock. This is fairly similar to the FFS superblock, except it again uses a B+Tree to store the allocation group references. The only "odd" thing about the XFS superblock is that because of where it is in the partition (right at the beginning), you cannot install a boot block to an XFS partition. There is simply no room for it. So, if you use XFS, you must install boot blocks to the MBR.
Benefits of XFSSo, what does this design give XFS? Well, scalability is obvious. UFS scales well to around 1 TB, NTFS to around 2 TB and XFS to around 18-19 million TB (these numbers are theoretical, I am just talking about the design here). Small difference there. It is also a very fast file system, allowing reads of up to 7 GB/sec. One thing some might find disturbing about the file system, is delayed allocation. It does this for good reasons, but the delay means that while a crash will not destroy FS, it might result in significant loss of data that has not yet been written to the disc. [1]
ConclusionThat is XFS: a fast, modern and extremely scalable file system. If you are storing huge numbers of large files on gigantic discs, give XFS a try. Even if you aren’t you might want to try it out, as some people have gotten great results using it. Personally, I am waiting to decide on my favorite FS. Reiser4 and ZFS are still lurking out there and they will be covered next. Oh yes...I’m front-selling.
References
Written by Mad Penguin™ System Administrator, Narayan Newton |
