[RFD] xfs_fsr: Adding features for online defragmentation

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hello,

We are a team of three final year under-graduates from Pune Institute
of Computer Technology, India and were looking for a project in XFS 
with a development time of about 6 - 7 months. On contacting Dave, 
he suggested we work on the userspace utility, xfs_fsr.
 
xfs_fsr(online defragmenter):
A few possible enhancements to xfs_fsr are as follows:
 
1. explicit control of locality for groups of files.
2. ability to defragment only portions of files rather
than just whole files.
3. defragmenting free space.
4. enhanced/faster filesystem scanning.
 
Our understanding of the problem is as follows:
 
A) We identified the first thing needed, will be to find regions of 
fragmented free space. Dave directed us to the following prototype 
that he had written
 
B)  The next task will be to go about finding the owners of the blocks
that are causing the region of fragmented free space. It will require a
reverse block map to do this efficiently. Dave has said that he is likely
to implement it in the next couple of months, and we understand that
we will to need to define the interfaces needed to query the rmap from 
userspace.
 
C)  Once the fragmented free space, and the files that own the blocks 
that fragment the free space are found we will need to move them, and 
take care of the following:
(a) move just the blocks we need to.
(b) control exactly where we move them to.
 
We are reviewing the recent patches on the mailing list for IOC_MOV_DATA, 
which allows an arbitrary range of one file to be swapped with an arbitrary
range of another file. The core mechanism required by #2.
 
Hence after due consideration we are planning to execute
the 4 sub topics in the following order:
1. defragmenting free space.
2. enhanced/faster filesystem scanning.
3. ability to defragment only portions of files rather
than just whole files.
4. explicit control of locality for groups of files.
 
Comments regarding the same will be appreciated. We would also like to thank 
Dave for his inputs. We look forward to working with the community and hope to
make a valuable contribution.
 
Regards.
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs

[Index of Archives]     [Linux XFS Devel]     [Linux Filesystem Development]     [Filesystem Testing]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux