tag:blogger.com,1999:blog-2890503559617005376.post5955383265014292960..comments2018-05-22T04:19:47.153-07:00Comments on Confessions of a Linux Penguin: Part III: Enter KVMEric Lee Greenhttp://www.blogger.com/profile/12350104299041375832noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-2890503559617005376.post-76311187093866973982013-07-02T21:40:17.482-07:002013-07-02T21:40:17.482-07:00I would definitely like to re-test zfs now that it...I would definitely like to re-test zfs now that it has reached what I consider to be release status. That said, some of the behavior I noticed from it appears to be inherent in the way it interacts with the Linux block layer. I'm sympathetic to their plight, the block layer is a pig and has been quite obviously one for quite some time (disclaimer -- I was the last maintainer of the Intransa block layer shim that integrated between the Intransa RAID subsystem and the block layer), and the reason btrfs gets excellent performance is by basically abusing the block layer via functionality specifically added to the block layer in order to speed up/make more reliable btrfs. But zfs can't modify the block layer to do their thing, they have to use it as-is due to licensing. So it goes.Eric Lee Greenhttps://www.blogger.com/profile/12350104299041375832noreply@blogger.comtag:blogger.com,1999:blog-2890503559617005376.post-34304032857764544432013-07-02T10:23:24.898-07:002013-07-02T10:23:24.898-07:00I would ask that you repeat the xfs virtual machin...I would ask that you repeat the xfs virtual machine zfs test using a zvol from zfs on the host. I don't expect amazing results, but I'd like to see how it holds up.Anonymoushttps://www.blogger.com/profile/10314162715108169250noreply@blogger.com