This is a new feature of rbd layering, when reading an object from child, if not exist, the kernel rbd client will not only request parent for the object, but also write it to child, and the jobs are done in an asynchronous way. Therefore, the subsequent accesses on this object will hit child without bothering parent. This feature could avoid longer latency incurred during accessing parent, especially when child and parent are geographically isolated, and it also could potentially avoid overloading parent. The patches: https://github.com/ceph/ceph-client/pull/11 Min Chen (4): Rbd: add an option for copy-on-read Rbd: add a new request: rbd_copyup_request Rbd: helper functions to manipulate rbd_copyup_request Rbd: implement the copy-on-read logic drivers/block/rbd.c | 385 +++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 382 insertions(+), 3 deletions(-) -- 1.9.1 -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
