Page 4 of 6
Re: Changes-only restore
Posted: Fri Aug 30, 2024 11:41 pm
by Brian K
You can check the Write/Read speeds on the partitions you are using for backup/restore with ATTO Disk Benchmark. What are the best results for the source and target partitions?
For a 2 TB HD I get 151 write 166 read (MB/sec)
For a 480 GB SSD I get 420 560
For a 1 GB NVMe M.2 I get 2700 3500
Re: Changes-only restore
Posted: Wed Sep 25, 2024 3:20 am
by n8chavez
For a 2 TB HD I get 150 write 162 read
For a 1 GB Samsung 990 Pro NVMe M.2 I get 2800 3565
I'm at a loss. I thought maybe it could be because the metadata files I use are hidden in Windows and thus not being used. I unhid them, and then restored. Same speeds. It's like the changed-only restore is being ignored, that or the metadata used is being ignored. I'm not sure why.
If I do delta restores in Reflect and Hasleo, the restores happen in under 25 seconds. But with IFW they take 2+min. It's not a time issue. What bothers me is I can't figure out why.
Re: Changes-only restore
Posted: Thu Sep 26, 2024 6:58 am
by TeraByte Support
one thing is we don't rely on potentially unreliable things like the usn journal. May be okay for file based if you trust anything that makes any changes is updating it.
Re: Changes-only restore
Posted: Thu Sep 26, 2024 7:44 am
by n8chavez
Forgive me, but I don't know if I quite know what that is. From what I understand it's just a record of changes. But I can't think of anything I'd use that makes use of that.
Re: Changes-only restore
Posted: Thu Sep 26, 2024 10:07 pm
by Brian K
n8chavez has a single NVMe M.2 drive and a HD. He has tested IFW restores with the backup image on the HD and the backup image on the NVMe. The times are much the same. In my computer, a restore with the image on the single NVMe drive would have been 8 times faster than with the image on a HD. Is there an explanation?
Re: Changes-only restore
Posted: Thu Sep 26, 2024 10:21 pm
by n8chavez
So, I just did some interesting tests. I created complete system images with IFW and another, one right after the other. Then I booted into my uefi partition and restored the other image first. Obviously, there is no data change between the disk and the image. The other restore took 15 seconds.
https://www.wilderssecurity.com/attachm ... ng.278568/
I then restored the IFW image, again, there is no data change between the disk and the image. That took 2:02. Very different than other's 15 sec. Both were supposed to be delta restores.
https://www.wilderssecurity.com/attachm ... ng.278569/
Then I erased the partitions, as a test. Then I rebooted and launched my UFD, reimaged with IFW again. Everything went fine, everything restored. But it took 2:05. So it essentially took the same about of time to rewrite an image I know has to be a complete restore (no partitions) and a delta restore? Something is wrong here. Changed only restore option is being ignored.
Re: Changes-only restore
Posted: Thu Sep 26, 2024 10:37 pm
by Brian K
n8chavez doesn't have a Microsoft Reserved Partition on that NVMe drive. Is that significant?
Re: Changes-only restore
Posted: Thu Sep 26, 2024 11:01 pm
by n8chavez
Everything runs fine. I guess the only reason I would recreate it is, is not having the MSR preventing delta restores? I couldn't imagine that being the case. But I prove to myself daily that I'm an idiot, so who knows.
Re: Changes-only restore
Posted: Thu Sep 26, 2024 11:25 pm
by TeraByte Support
It doesn't use the usn jorunal to look for changes .. it checks all the metadata of the files (ntfs). if you think it's writing the same, you could use something like filemon / process monitor to see (filter for writes). of course the metadata hash file has to exist for the restore to use it or it falls back to normal restore.
I meant diskmon
Re: Changes-only restore
Posted: Thu Sep 26, 2024 11:38 pm
by n8chavez
My metadata files do exist. Is there something that could be preventing them from being used?
https://www.wilderssecurity.com/attachm ... ng.278572/