On 12/29/2012 03:45 PM, Richard Ryniker wrote: >> How does having two PVs on the same spindle increase latency? > > Longer seeks - from one PV to the other - instead of two writes into the > same PV. > Or the opposite depending on where the data is placed. One PV/VG: Create a 200G guest, make a (maybe temporary) backup of it to /root/guest1-backup.img and the create a second 200G guest. Now you have two guests with a 200G gap between them which means if both are running and do disk i/o the seeks will have to jump this gap all the time. One system PV/VG and one separate guest PV/VG: Same scenario as above except now the backup image will end up in the system VG and the guests will be placed next to each other in the guest VG which will result in faster seek times. The only way to guarantee seek times is to control both the exact placement of the data and the access profile which defeats the purpose of using a logical volume manager. Using logical partitioning while at the same time making assumptions about the physical access properties isn't a good idea and really goes against the abstraction lvm tries to provide. Regards, Dennis -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
