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

Re: XFS vs. ext3



On Wed, Feb 26, 2003 at 02:20:30PM +0000,
Nigel Metheringham wrote:

> Would you like to elaborate on what these tests were doing,
> and what exactly a dropped frame is in this context (for those
> that aren't familiar with this sort of work).
> 
> I assume you basically have 4 boxes writing lots of data to a
> server over Samba shares.
> Are they writing individual files, or components within larger
> files (ie is (file level) locking involved in this)?

They are writing individual files.  Each client writes fifty
files sequentially; open one file, write to it, close the file,
open the next file, write to it, close it, and so on.  Different
processes are never attempting to write to the same file; client
1 gets frames 0-49, client 2 gets frames 50-99, etc.  So at the
same time, client 1 will be writing file 0, client 2 will be
writing file 50, client 3 will be writing file 100, and client 4
will be writing file 150.

No client waits for the other clients to finish their frame
before moving on; near the end of a test run, you might see,
say, client 1 writing file 48, client 2 writing file 95, client
3 writing file 149, and client 4 writing file 198.

> Is a dropped frame equivalent to a missing file, or a corrupt
> file or a hole within a file?

A dropped frame is equivalent to a partially written file.

The application appears to be encountering an error or a
disconnect partway through writing a frame/file.  It then
immediately moves on to writing the next frame/file in the
sequence, leaving a partially written frame/file on the server.

Andrew Klaassen





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