| [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] |
On Thu, 2003-05-08 at 15:34, Alan Cox wrote: > On Gwe, 2003-05-09 at 00:15, Joe DiMartino wrote: > > It looks like removes are the biggest challenge. I don't know why > > they are 40x slower than ext3. It also looks like read caching > > isn't working properly as the memexp was 11.72x slower. > > Deletion is an expensive business in a file system, especially when > you have multiple nodes who might be touching a file. I seem to remember > opengfs delete is also synchronous (that is unlink() returns after the > deletion hit disk) > This is an excellent point. I failed to mention in my (already lengthy) description that the times included a sync after each step. In other words: untar = time tar x + time sync tar = time tar c + time sync rm = time rm -rf + time sync So I was hoping that even in the ext3 case, the sync would force the unlink() to disk, thus leveling the field. But it's possible that OpenGFS is getting charged for a synchronous I/O on every unlink() and that ext3 gets to batch it somehow. ------------------------------------------------------- Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara The only event dedicated to issues related to Linux enterprise solutions www.enterpriselinuxforum.com _______________________________________________ Opengfs-devel mailing list Opengfs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/opengfs-devel