> We will to have add tests cases for checking whether the filesystem
> degrades gracefully when it is being filled up but design is not the
> scope of the test cases. We would trust the experts, in this case, the
> filesystem developers to take care of that.

To take care of the design, yes, but test case, no.  I reckon this is
something I can go ahead and hash out without waiting for feature freeze.


I'm thinking something along the lines of (very rough, but illustrates the point):

Install
Run benchmark speed test
put  in a bunch of data
Run benchmark speed test (read/write, etc.) when half full
Repeat at 3/4
Repeat at nearly full.
Maybe compare against the same partition/drive with ext4
running the same test with (preferably) the same data?

The question is, how much degradation is acceptable?  I'm thinking the
standard to measure against is ext4 since that is current.  The issue is,
btrfs doesn't even meet ext4 standards now.  It's getting better, but isn't
there, yet.  This is especially so with large files such as those used with
virtual machines.

John.