I've been wrestling with a problem that really has me baffled. I'm running W8.1 Pro as a host Hyper-V. On the host there's 10 4t drives with Drivepool and Scanner (currently disabled). There's one guest OS which is WHS2011 running 4 2T drives on native motherboard ports. They are offline in the host, and then allocated to WHS in the guest configuration, and become a part of a second Drivepool running on WHS. Like on the host, scanner is disabled.
I'm using Drivepool Beta *.432 presently.
The host is a quad core i7 with hyper threading disabled, and two cores allocated to the WHS guest. It runs around 3Ghz clock and very low CPU utilization on both host and guest. There's plenty of memory.
Everything 'works', no errors no blowups, but during backups for other computers in the house, WHS takes prizes for slowness. It's probably 8x slower than my ancient WHSv1 system. The 4 drive pool is the target for Client backups.
I've been studying just what is happening and noticed there are only occasional bursts of data being sent from the machine being backed up to WHS. They are a couple of minutes in length, and typically run in the area of 900Mbits over the net (seen at WHS). But they stop for sometimes as much as ten minutes before another burst. Since these machines are being backed up for the first time (on WHS2011) all blocks should be 'unbacked' and subject to being sent.
The standard data blocks written to the drivepool are 4GB in size (seen in the directory), but there are occasional Global data files that could be 17GB or more. But during the time of no data from the backed up machine, hovering over the Disk Performance indicator, the drive which is showing active solid, is also showing both read and write activity of the same speed, and it is very very slow. Like around 15-20MBytes/second. Imagine how long it takes to copy a 4-20GB file at 20MB/second! As soon as that process is done, the data flow from the backed up computer immediately starts up again.
Sometimes the file indicated for the drive enduring the slow transfer is *.new, *.tmp or *.dat. It appears to be copying these files to the same physical drive. It's as if Drivepool is converting what should be a file move or rename to a copy then delete. I can't see any reason for all these copies going on. When there is no drive running solid, and data is streaming in from the backed-up machine the drive activity data is upwards of 100MB/sec, so I don't think it's the drives themselves, though older ones would certainly be much slower than the latest generation.
Is this a WHS issue, or a DrivePool issue? It makes using WHS2011 downright painful. I'm on my third day of backing up just one machine and not even half way done! I don't have problems writing my own data from a remote host to the WHS pool, this strange behavior happens only during backups and only during these copy processes.
What to do? Oh, I tried without duplication and without balancing, but nothing made any difference. Whether a balancing is in effect at the moment doesn't seem to bother anything (make it better or worse). I've removed the drives from port adapters and connected all of them directly to the motherboard 6G SATA ports. That improved it slightly, but not significantly.
Question
bblue
I've been wrestling with a problem that really has me baffled. I'm running W8.1 Pro as a host Hyper-V. On the host there's 10 4t drives with Drivepool and Scanner (currently disabled). There's one guest OS which is WHS2011 running 4 2T drives on native motherboard ports. They are offline in the host, and then allocated to WHS in the guest configuration, and become a part of a second Drivepool running on WHS. Like on the host, scanner is disabled.
I'm using Drivepool Beta *.432 presently.
The host is a quad core i7 with hyper threading disabled, and two cores allocated to the WHS guest. It runs around 3Ghz clock and very low CPU utilization on both host and guest. There's plenty of memory.
Everything 'works', no errors no blowups, but during backups for other computers in the house, WHS takes prizes for slowness. It's probably 8x slower than my ancient WHSv1 system. The 4 drive pool is the target for Client backups.
I've been studying just what is happening and noticed there are only occasional bursts of data being sent from the machine being backed up to WHS. They are a couple of minutes in length, and typically run in the area of 900Mbits over the net (seen at WHS). But they stop for sometimes as much as ten minutes before another burst. Since these machines are being backed up for the first time (on WHS2011) all blocks should be 'unbacked' and subject to being sent.
The standard data blocks written to the drivepool are 4GB in size (seen in the directory), but there are occasional Global data files that could be 17GB or more. But during the time of no data from the backed up machine, hovering over the Disk Performance indicator, the drive which is showing active solid, is also showing both read and write activity of the same speed, and it is very very slow. Like around 15-20MBytes/second. Imagine how long it takes to copy a 4-20GB file at 20MB/second! As soon as that process is done, the data flow from the backed up computer immediately starts up again.
Sometimes the file indicated for the drive enduring the slow transfer is *.new, *.tmp or *.dat. It appears to be copying these files to the same physical drive. It's as if Drivepool is converting what should be a file move or rename to a copy then delete. I can't see any reason for all these copies going on. When there is no drive running solid, and data is streaming in from the backed-up machine the drive activity data is upwards of 100MB/sec, so I don't think it's the drives themselves, though older ones would certainly be much slower than the latest generation.
Is this a WHS issue, or a DrivePool issue? It makes using WHS2011 downright painful. I'm on my third day of backing up just one machine and not even half way done! I don't have problems writing my own data from a remote host to the WHS pool, this strange behavior happens only during backups and only during these copy processes.
What to do? Oh, I tried without duplication and without balancing, but nothing made any difference. Whether a balancing is in effect at the moment doesn't seem to bother anything (make it better or worse). I've removed the drives from port adapters and connected all of them directly to the motherboard 6G SATA ports. That improved it slightly, but not significantly.
Anyone with thoughts or suggestions?
--Bill
Link to comment
Share on other sites
24 answers to this question
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.