Shirt Pocket Discussions

Shirt Pocket Discussions (https://www.shirt-pocket.com/forums/index.php)
-   General (https://www.shirt-pocket.com/forums/forumdisplay.php?f=6)
-   -   Backup to HFS+ or APFS (https://www.shirt-pocket.com/forums/showthread.php?t=7124)

Zarkov 04-17-2018 04:55 PM

Backup to HFS+ or APFS
 
Hi Dave, et al-

I presently backup my SSD-based iMac to an external, directly connected HFS+ formatted SSD drive. I understand I can backup to the external SSD if it is formatted as either HFS+ or APFS. Are there advantages or disadvantages to either approach? Thanks!

dnanian 04-17-2018 05:27 PM

I'd format it as APFS if that's an option. See http://www.shirt-pocket.com/blog for a bunch of discussion of the various advantages of APFS in this situation...

hand123 01-09-2019 01:01 PM

Hello,
I've read somewhere that it is not advised to format an external drive with APFS. Is that true?

I would like to make my SuperDuper clone of my internal Mac mini 2018 SSD APFS to an external USBC SSD as similar as possible. Therefore I would like to format the external with APFS is this is ok.

Thanks for any additional feedback and guidance.

dnanian 01-09-2019 02:31 PM

Definitely format it as APFS.

d.r.h. 02-06-2019 06:03 AM

Some questions about backups on APFS formatted SSDs. Background: My main computer is a 3.6 GHz Intel Core i7 iMac with 32GB RAM and a 1TB SSD. I used to backup to two HFS+ formatted drives (a 750GB WD Scorpio Black, 7200 RPM - Daily, and a 500GB Hitachi, 5400 RPM - weekly). Both connected via USB3. I use a similar 500GB Hitachi for periodic backups on my Macbook pro (which has a 500GB internal SSD). I recently replaced the Hitachis with Otherworld Computing Electra 6G SSDs (a 1TB for the iMac and a 500GB for the MBP). Formatted both of these new SSDs APFS. After formatting, ran SD to do an erase and copy to the SSDs.

What confuses me is that my daily iMac smart backup to the new SSD ( going from SSD to SSD) is really not that much faster than to the WD Scorpio (maybe a minute faster ... in the 6 minute vs 7 minute range). I thought it would be much faster. This is not a big problem, as I would have still bought the SSDS. However, would really like to understand this.

Prior to doing a comparison test, I did a smart update on both so that they each would have a current clone of my iMac. I then used the computer for a while and then redid a smart update on each.

I did immediately notice one thing that was different when the backups started. When smart updating to the SSD, it took 30 seconds to create and mount a snapshot of the iMac ... where creating and mounting a snapshot smart updating the Scorpio is pretty much instantaneous. Here are snippets from my logs (backed up Scorpio first then immediately backed up the SSD):

Scorpio:

07:00:30 PM | Info | ...ACTION: Preparing NachalnikiMac snapshot
| 07:00:30 PM | Info | ......COMMAND => Creating Snapshot of NachalnikiMac
| 07:00:31 PM | Info | ......COMMAND => Mounting Snapshot for NachalnikiMac

SSD:

07:11:07 PM | Info | ...ACTION: Preparing NachalnikiMac snapshot
| 07:11:07 PM | Info | ......COMMAND => Creating Snapshot of NachalnikiMac
| 07:11:37 PM | Info | ......COMMAND => Mounting Snapshot for NachalnikiMac

I also ran a smart update on the Macbook Pro to check the time it took to create and mount a snapshot to the new SSD I use for that computer. It took 61 seconds:

10:52:53 AM | Info | ...ACTION: Preparing Nachalnik_MBP snapshot
| 10:52:53 AM | Info | ......COMMAND => Creating Snapshot of Nachalnik_MBP
| 10:53:54 AM | Info | ......COMMAND => Mounting Snapshot for Nachalnik_MBP


Here are the pertinent snippets from the logs showing the actual data evaluated and copied/cloned info:

Scorpio:

07:06:03 PM | Info | Evaluated 1077918 items occupying 151.85 GB (182758 directories, 875744 files, 19416 symlinks)
| 07:06:03 PM | Info | Copied 5341 items totaling 0.56 GB (180767 directories, 5325 files, 16 symlinks)
| 07:06:03 PM | Info | Cloned 146.05 GB of data in 331 seconds at an effective transfer rate of 441.23

SSD:

07:17:09 PM | Info | Evaluated 1077929 items occupying 151.85 GB (182763 directories, 875750 files, 19416 symlinks)
| 07:17:09 PM | Info | Copied 5222 items totaling 0.56 GB (710 directories, 5206 files, 16 symlinks)
| 07:17:09 PM | Info | Cloned 146.05 GB of data in 331 seconds at an effective transfer rate of 441.24

The time actually performing cloning on each and the effective transfer rate is just about the same, 331 seconds at around a 441 MB/s effective transfer rate.

Looking at the complete logs shows a very large number of actions performed by SD when copying to the SSD vs Scorpio which I assume are because of the format difference between the two. I assume this (along with the increased time to create and mount the snapshot) attributes to the overall amount of time SD takes to create the backup to the SSD. Also, not quite sure I understand the difference in the directories copied data info on the SSD vs Scorpio as recorded above:

Scorpio:

Copied 5341 items totaling 0.56 GB (180767 directories, 5325 files, 16 symlinks)

SSD:

Copied 5222 items totaling 0.56 GB (710 directories, 5206 files, 16 symlinks)

While this is all probably normal ... hoping you have some info that would help me understand why the overall time it takes to back up to a much faster SSD formatted to APFS is not that much different than to a somewhat fast regular HD formatted HFS+

Thanks.

dnanian 02-06-2019 09:10 AM

This is going to be a disappointing reply, Dennis, but - the fact is, Smart Updates involve reading a zillion teeny bits of information (metadata), comparing them, walking the directory trees, etc.

So there's a lot of "back and forth" between a lot of layers of the OS...and speeding up the drive only matters to a certain extent (when things are generally not being copied).

Since a lot of that information is directory information, and the directory is aggressively cached, speeding up the drive won't make a huge difference.

Similarly, copying, say, a giant file is faster than copying a million tiny files that add up to the same size...because the "bookkeeping" time ends up being a larger percentage of time (you know, a million calls to the OS vs. a smaller amount).

With regard to creating the snapshot, well, we ask the system to do it and it takes what it takes. You'll note that we're not CPU bound - we're mostly waiting for the OS to do things...

d.r.h. 02-06-2019 09:46 AM

Thanks Dave ... I guess it really is what it is! It just came as a surprise to me (especially when expecting different results). Biggest surprise was the time difference in creating a snapshot when connected to an HFS+ conventional drive vs an APFS SSD. Still can't wrap my head around that. Thanks again.

dnanian 02-06-2019 10:20 AM

Snapshots aren't created on HFS+ volumes (they don't support snapshots). With an APFS SSD, you're getting a snapshot on the source and the destination.

d.r.h. 02-07-2019 05:12 AM

Quote:

Originally Posted by dnanian (Post 34416)
Snapshots aren't created on HFS+ volumes

Thanks for clearing that up. I find my old mind gets confused quite easily nowadays! I would not have know that just looking at the logs, since the log for the HFS+ formatted drive kinda showed one being created, though in a fraction of a second!

dnanian 02-07-2019 08:54 AM

Much like what happens when we get to the "make bootable" step with a drive that doesn't have an OS on it, a step succeeds (quickly) if it determines that there's nothing to actually do.

Errors happen when there's something to do and it fails.


All times are GMT -4. The time now is 07:41 PM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2024, vBulletin Solutions, Inc.