Is Time Machine not a very good program for backing up the MBPs?

Discussion in 'macOS' started by hajime, Apr 9, 2010.

  1. hajime macrumors 68040

    Joined:
    Jul 23, 2007
    #1
    Hello. I have the motherboard of my SR MBP replaced recently. I followed the instructions on http://www.macosxhints.com/article.phpstory=20080128003716101 but it did not work out. I tried to create a completely new backup but the 1G HD does not have enough disk space anymore. It seems that it is better to use SuperDuper to backup the MBPs for the following reasons:

    1. If you have a firewire drive, you can restore the content of the entire disk easily.

    2. It works even the motherboard has been replaced. This is good especially for those who have the SR MBP with defective GPU.

    3. I am not sure how it works but it seems that when the Time Machine backup drive is almost full, TM asks the user for permission to delete
    older backups. Some of the important stuffs may be deleted accidently. Is this possible?

    Supposing that I have backup_on_Jan_2009, backup_on_Feb_2009, ..., backup_on_April_2010. As far as I understand, TM does incremental
    backup. Assuming that I have added and deleted folders throughout the year. If I allow TM to delete earlier backups, will some of the files be
    deleted unexpectedly?

    Any comments/suggestions are appreciated.
     
  2. Pax macrumors 6502a

    Joined:
    Dec 12, 2003
    #2
    I personally like SuperDuper for the reasons you give. When I use SuperDuper I know I have a full, perfect, bootable backup. It has saved me once when my hard disk went wrong and I was quite busy with work. I just plugged in my SuperDuper drive and carried on working.

    I have set up an auto-backup routine using Folder Actions and a script so when I plug in my SuperDuper drive it backs up, ejects the drive, quits SuperDuper and sleeps the computer. So every evening at beer time I just plug in the Magsafe, plug in the SuperDuper drive and leave everything.

    I guess my big problem with Time Machine is that I don't understand exactly how it works. So I don't have as much confidence in it as SuperDuper. If I ever start to worry about SuperDuper, I simply reboot my computer from the external drive, if it works OK I am confident that I am safe.

    I am sure Time Machine is excellent, I have just never learned how to use it. I can imagine it would be very good for someone with a workflow that involved many file, eg photo editing. TM would track all the versions of all the photos and would back up throughout the working day. My workflow is mostly Word documents. I might only change 4 or 5 files in a working day, not hundreds. I am vulnerable until I do my end-of-day backup. If I used TM I would not have that vulnerability.
     
  3. emptyCup macrumors 65816

    emptyCup

    Joined:
    Jan 5, 2005
    #3
    I use both and both are excellent for different things.

    SuperDuper is best for catastrophic HD failure requiring restoring the whole disk. Along the same lines, it is excellent for rolling back an OS update you don't like. The problem is "How many backups can you have?". I backup once a week. Most average users, if they backup at all, can only afford a single backup drive.

    Time Machine is excellent for cases where a specific instance of a file is missing or corrupt. If I decide I don't like the changes I made to a file and want to restore the version I had two weeks (or two years) ago, I can.

    Yes, you can restore a whole disk from TM, but it is slow. And, yes, you can restore specific files from SuperDuper, as long as you haven't overwritten the version you want with the latest backup.

    If you are not backing up at all, Time Machine is painless and automatic (after the initial backup). Ideal would be a HD large enough for both SuperDuper and Time machine partitions. Otherwise, pick the situation that best suits your needs.
     
  4. hajime thread starter macrumors 68040

    Joined:
    Jul 23, 2007
    #4
    Thanks for the suggestions.

    Having both SuperDuper and TM backups on the same HD may be a bit risky. Maybe the best way is to get two external drives. One for SuperDuper and the other for TM backups?

    Assuming that the internal drive of the MBP is X GB, what is the recommended size of the external backup drive? From your post, it looks like even I use a drive for backing up via SuperDuper, I can save other files (perhaps different versions of the backups created by SuperDuper) into the same drive. Am I right? If so, how SuperDuper knows which version to restore or boot from when the internal drive fails? Is Firewire the recommended interface to use with SuperDuper? Thanks.
     
  5. Pax macrumors 6502a

    Joined:
    Dec 12, 2003
    #5
    Answering some of your questions:

    SuperDuper works equally well on Firewire and USB. I had a MacBook and used Firewire. I gave that to my girlfriend and got a MacBook Pro, and now I use USB. I deliberately changed to USB because it looks like Apple is deprecating Firewire, at least on the cheaper computers. I believe Firewire may be a little faster, but not much.

    I only have about 70 gig of data, not very much.

    SuperDuper simply "clones" your entire hard disk every time you run it. The free version copies every single file. For my 70 gig that takes more than an hour. If you pay for SuperDuper it unlocks an incremental backup mode which only copies what it needs to to make the two disks identical. For me that takes 10 - 15 minutes each time.

    The key point is that every time you run SuperDuper it makes the internal and backup disks identical (apart from some cache files).

    If your internal disk fails, you have an identical clone on your backup disk. If you like you can make as many backups as you like, at different times.

    Your question about how SuperDuper knows what to restore is not really relevant. Each SuperDuper backup contains only one copy of each file - the copy that was on your internal drive at the time of the backup. There is no file "history" like there is with TM. If your internal drive fails you can only restore the file back to exactly how it was last time you ran SuperDuper. That is SuperDuper's big disadvantage compared with TM.

    You can run multiple SuperDuper backups. As the previous poster said, SuperDuper is great for rolling back a software update that didn't work. My backup drive is 120 gig. When I had <60 gig on my Mac I was able to have 2 partitions on my backup drive. On the first I left a SuperDuper backup made before a major update (eg 10.6.2 -> 10.6.3). I left this backup untouched for weeks or months. On the second I put my daily 10.6.3 backups. That way if 10.6.3 caused me a problem I could restore the system to 10.6.2 from the first backup, and the data from the second. But now I have >60 gig data I can't do this unless I buy a bigger backup drive!

    I think if I need to buy another backup drive, I will buy a BIG one, with Y gig where Y ~ 5X. I will partition it so that one partition is X gig, and I will use that for SuperDuper. The remaining Y-X gig I will use for TM. I will use my SuperDuper script to do a daily "clone" backup, and rely on TM for the rest.

    In fact I really should do that.... 1 TB external is only UKpounds 50......
     
  6. iThinkergoiMac macrumors 68030

    Joined:
    Jan 20, 2010
    Location:
    Terra
    #6
    Huh?! FW400 isn't much faster than USB2, but FW800 is more than 2x as fast as USB2. I'm hoping you're talking about FW400, but it doesn't apply to most Macs since all the new ones (and many older ones) have FW800. Heck, my 6 yr old PBG4 has FW800.

    When I moved my TM backup from a FW400 IDE drive to a FW800 SATA drive the difference was insane. Backups take less than half the time they used to.
     
  7. Garsun macrumors regular

    Garsun

    Joined:
    Oct 20, 2009
    #7
    Point #3 is not an issue

    Put your mind at ease. Even after the backup_on_Jan_2009 has been deleted, all the files existing on your MBP on the date of the "backup_on_Feb_2009" backup will still exist.

    Every TM backup presents as a full backup not as a incremental backup, even though it only has to move the new stuff over once an hour.....

    Its one of those cases of I know how it works but when I hear myself describe it it sounds ... well bad.

    If I can come up with, or find a description of how it works that does not sound confusing I'll post it here.
     
  8. CarlJ macrumors 68020

    CarlJ

    Joined:
    Feb 23, 2004
    Location:
    San Diego, CA, USA
    #8
    How Time Machine works...

    I wanted to write this as a separate post rather than burying it in the middle of my (next) reply, since this one is rather convoluted...

    First, it helps to understand a little bit about Unix filesystem structure (this bit may be confusing, but helps later, and yes, I'm oversimplifying things somewhat; if you already understand inodes and hard links, skip this list of bullet points):
    • On a Unix filesystem, the blocks that make up a file are listed-in/pointed-to-by a structure called an inode. Most of the "metadata" that you associate with a file (its owner, permissions, timestamps, etc.) are stored in this inode, but the file's name is not stored here (more on that in a moment).
    • There's a table (the "inode table") that holds all these inodes. It's just a big array. inodes are commonly referred to by their "inode number", which is just an index into this array.
    • Every entry in a directory/folder, is really just a pointer to the corresponding inode. A directory is just a special type of file that lists filenames and their corresponding inode numbers.
    • Most files show up only one place in the filesystem (i.e. they're listed in only one directory somewhere), but files can just as easily appear in more than one place (not copies of files, the exact same file appearing in two or more places at once). This is completely normal, just not used that often. But here's one example you can see in the Terminal, if you're interested:
      Code:
      $ ls -li /bin/*sh
      10154076 -rwxr-xr-x  1 root  wheel  1346544 Feb  4 16:41 /bin/bash
      10119754 -rwxr-xr-x  2 root  wheel   767200 Feb 10 21:54 /bin/csh
       4970320 -r-xr-xr-x  1 root  wheel  2186880 May 18  2009 /bin/ksh
      10154122 -r-xr-xr-x  1 root  wheel  1346624 Feb  4 16:43 /bin/sh
      10119754 -rwxr-xr-x  2 root  wheel   767200 Feb 10 21:54 /bin/tcsh
       4971547 -rwxr-xr-x  1 root  wheel  1597200 May 10  2009 /bin/zsh
      $
      the "i" in "-li" tells ls to display file's inode numbers; the first column in the output is the inode number (the second is the file's permissions) and the third column is the file's link count -- the number of places it shows up in the filesystem (for completeness sake, the remaining columns are the file's owner, group, size in bytes, modification time, and name). You'll note that most files have a link count of 1, but csh and tcsh have a link count of 2, and their inode numbers are the same. This is because they are two directory entries for the same file. This is different from aliases or symlinks, where one entry is the "real" file and the other is a pointer to the "real" file. These are two first-class equally important names for the same file.
    • Making a new link to an existing file is called making a "hard link" (to distinguish these from symbolic links, another vaguely related mechanism available on Unix). (Please please don't play around with this if you don't fully understand it, you can cause yourself all sorts of grief).
    • When a file gets deleted, the name is erased from the directory where you're deleting the file, and the link count in the corresponding inode is decremented by one. In most cases, this means the link count is going from 1 to 0. And when the link count reaches 0, the OS erases the corresponding inode and frees up the file's disk blocks. But if a file has multiple links (appears with multiple names in the filesystem's directory structure), its link count might be decrementing, say, from 5 or 4, or 2 to 1; in those cases, the file still exists, attached to the filesystem's directory structure at those other names, so the inode and blocks stay intact.
    • A new feature, which Apple added to Mac OS X added for Time Machine, and doesn't exist in previous Unix filesystems is, it's also possible to make a hard link to a directory, rather than just to a file (this is potentially dangerous, as you can create broken directory structures with loops in them, so it's carefully controlled).
    Okay, backstory complete. Now, explaining Time Machine is easier.

    Time Machine makes full backups of your drive (every time), but it uses some ingenious tricks to make these backups nearly as small and quick as incremental backups. A new backup can be thought of as: "just like the previous backup, plus any files added/changed since the previous backup, minus any files deleted since the previous backup."

    So, Time Machine asks the operating system (Mac OS X) to keep track of every file that is added, modified, or deleted. Mac OS X is very good at this.

    Then, when it's time to do a backup, Time Machine says, "okay, what's everything that changed since last time" (it'll either instantly get that list that the OS has kept, or in rare cases it'll get "sorry, don't have it, go look at everything and figure it out yourself", which takes much extra time).

    Time Machine then makes a new empty directory for the backup ("/Volumes/(BackupDrive)/Backups.backupdb/(MachineName)/(yyyy-mm-dd-HHMMSS)", fill in the appropriate bits and that last part is a timestamp), and runs down the directory tree of the previous backup, making a copy of it into the new backup directory, except it does a couple of special things:
    1. when it gets to a file that's in that list of files that got changed or added, it copies that file from your "real" drive into the backup; if it gets to a file that has been deleted, it simply doesn't put anything in the new backup for it; and,
    2. when it gets to a file or folder that it knows hasn't changed, it simply makes a hard link to the corresponding file or folder in the previous backup.
    Think about it: if nothing has changed in, say, /System, since the last backup, Time Machine simply makes a hard link from this backup's /System to the previous backup's /System, and goes on to the next thing. So this can all go very quickly, yet you end up with a new backup directory that contains an accurate copy of the entire filesystem.

    This means that files that get put into your initial backup, and then never change (which accounts for most of Mac OS X itself), will get written onto the backup disk once, and then hard linked into every backup after that -- each of those files may be linked into hundreds of backups.

    And when any given backup is deleted, Time Machine simply runs down the corresponding directory... any file or directory in the backup with a link count of 1 gets (truly) deleted, but any file or directory in the backup with a link count higher than 1, simply gets its link counter decremented, and continues to live on in the other backups that link to it.

    It's really a marvelous backup design, overall -- tricky to get right, but Apple seems to have pulled it off quite well.
     
  9. gnasher729 macrumors P6

    gnasher729

    Joined:
    Nov 25, 2005
    #9
    With your backups of Jan 2009, Feb 2009, and so on, you could restore your hard drive exactly to the state that it had in January 2009. Or if you deleted a file in February 2009, and you just noticed, you can get that file back from the Jan 2009 backup. Or if you changed a file many times over the last year, you can get a copy exactly as it was in January 2009.

    When the Jan 2009 backup is deleted, all the files in the February backup, March backup and so on are still there. But if you deleted a file near the end of January 2009, so it's not in the February backup, that file is now gone. And you can't get a file back exactly in the state it was in Jan 2009, only February, March and so on.

    Time Machine really does two things: It gives you a backup with the latest state, and a "time machine" that can give you files from the past. When the Jan. 2009 backup is deleted, you use the ability to go back to Jan. 2009.

    Time Machine makes an exact copy of your hard drive every time, but it uses two clever tricks to make it fast and to make it take little space: First, say you have one million files on your hard drive. 999,990 of those are exactly the same as an hour ago. Those 999,990 files are not copied to your backup, but Time Machine just takes a note that the file is the same as the one from the previous backup. But 999,990 notes would still be a lot of work. Now these files are in say 100,000 folders. 99,995 of those folders are the same as in the previous backup. So Time Machine just makes a note that a folder is the same as in the previous backup. Say your application folder is unchanged. To backup the complete application folder, with all its subfolders, all that Time Machine does is make a single note "Application Folder is the same as before". Same for all unchanged folders. In the end, only files or directories that have been changed or newly created will be copied.

    The second trick: The operating system keeps track of all directories that you are changing all the time. So when you start a backup, it knows where to look for differences. If you never changed the application folder, Time Machine won't find any note that the application folder was changed, so it just makes a note that it is unchanged without having to check if there were any changes.

    This makes the hourly backups very fast if your hard drive is permanently connected to the computer, because usually you don't have many changes. You won't even notice it when the backup happens except if you look at the icons in the menubar.
     
  10. CarlJ macrumors 68020

    CarlJ

    Joined:
    Feb 23, 2004
    Location:
    San Diego, CA, USA
    #10
    Time Machine does (more or less) do backups incrementally, but what you end up with is really more like a whole series of full machine backups (though they aren't ones that you can boot from, the way you can with SuperDuper's disk clones).

    And I have had occasion to restore my whole machine from a Time Machine backup, and was happy with the results (it restored all of my folders and applications and all of Mac OS X just fine; it left off some log files that are mostly of interest to Unix nerds, that I would have liked to have, but that wasn't a major problem).

    (By the way, rebooting a machine on a SuperDuper backup is reassuring, and a nice sanity check that things are probably mostly okay, but it is absolutely not a thorough test of a backup -- unless you then proceed to carefully examine every single file on the system -- it's more like turning your key and having your car's engine start, and then thinking that, therefore, your car is in perfect condition -- of course many things could still be wrong.)

    (And FWIW, I'm quite interested in the field of backup technology, was in charge of the backups at a company I used to work for, and engineered my own backup systems for their use, and we used those systems on several occasions to save the company's butt.)

    I run Time Machine and SuperDuper on the same drive with no problems at all, and I don't believe there's any meaningful risk in doing this. SuperDuper was designed to handle precisely this situation, and Time Machine doesn't venture outside of the Backups.backupdb folder in the root directory of the hard drive, so they don't step on each other's toes. My work machine has its (500GB FW800) backup drive attached all the time, so it does Time Machine backups every hour, and I refresh the one SuperDuper clone on it every few weeks (and I taught SuperDuper to ignore a number of large directories that I know Time Machine is handling, like my iTunes collection, to keep the size of the SuperDuper clone down to leave more space for Time Machine).

    Time Machine does delete backups as space on the backup drive runs low, but it uses a fairly smart algorithm (quoting Apple): "Time Machine saves the hourly backups for the past 24 hours, daily backups for the past month, and weekly backups for everything older than a month." And when it deletes a given backup, the only files you can "lose" are those that weren't around before that backup and weren't around after that backup.

    Time Machine has several huge wins over most other backup methods... First is, it's so d*mn easy to use that you actually will use it. When a drive fails, the Time Machine backup that you actually did make, and keep up to date, just by leaving the backup drive plugged in and turned on, is way better than any other backup that you had been meaning to do (or meaning to run again), but hadn't gotten around to doing. Good intentions don't count for anything when you need to reload the drive. :)

    Second win is, it tracks changes to everything on the hard drive (minus, as noted above, some things like a handful of log files that most people will never notice). This is a huge win over any backup plan that involves, "I back up the folders where my work is," because if you put a file somewhere else, you might fail to add that file to your backup strategy, but Time Machine will always catch it, since it looks everywhere. "Oh yeah, I guess I forgot to add that folder to my backup" also doesn't count for anything when you need to reload the drive. :)

    SuperDuper is much faster if you have to reload your entire drive, but then, on the rare occasion that I lose an entire drive, I don't care whether it takes one hour or four to recover, I'm just extremely thankful that I have backups and can recover the drive at all. If you're losing drives at a rate where that 1-hour-vs-4-hours deal really starts eating into your work week, perhaps you should be taking a hard look at why your drives are failing so often (a good first shot would be, how clean is the power you're feeding your computer? Do you need a UPS or line conditioner? Voltage spikes or low voltage conditions can do nasty things to equipment). I'm much more likely to want to quickly check what a given file looked like, say, three weeks ago, vs. two weeks ago, vs. yesterday, vs. today -- and that's something Time Machine can handle quite easily, while SuperDuper can't, unless you've got a whole stack of backup drives and you're extremely diligent about rotating backups through them.

    FWIW, FireWire is a better interface than USB for hard drives (FireWire was designed to handle things like hard drives, while USB was designed more for things like mice), though USB works well enough these days. Given the choice, I'd go with FireWire (on paper, USB2, at 480Mbps, looks faster than FW400, at 400Mbps, but FireWire can actually attain those speeds, while USB usually can't, plus USB puts a much heavier load on the CPU to do that same transfer). And FW800 eats USB2 for lunch :)

    This article is old, but still a very enlightening read, about how difficult it is to really get backups right: Mac Backup Software Harmful (and to give away the punchline, before Time Machine came out, the only software that really got backups right was SuperDuper).

    As far as drive space goes, at work, I've got a MBP with a 320 GB drive, with about 115 GB used, backing up with Time Machine to an always-on 500 GB external drive that's about 2/3rds full now, with backups going back to the start of the year. Previously, I had backups going back about a year and a half on the same drive, connected to my previous workstation, but I cannibalized most of those when my machine got replaced (I kept a few "key" backups of the older system around, freed up most of the drive for the new system to use). On one hand, the minimum space you need for Time Machine is, "a size equal to the amount of space you're actually using on your drive, plus some more", because otherwise Time Machine can't do anything useful. But you'll be happier with something more like "3 to 5 times what you're actually using", so, say a 300-500 GB drive if you're currently using 100 GB. That'd give Time Machine lots of room for copies of things, so you'll end up with backups going much further back in time. Then again, 1 TB drives aren't that expensive these days... And if you're serious about backups, and can afford it, what I'd really like to have at home for Time Machine backups would be a RAIDed NAS box from QNAP or ReadyNAS, but they are spendy.

    The way Time Machine works, it effectively stores only changed files for each new backup, so you can keep many MANY backups on the same drive, if most of your files aren't changing most of the time (e.g. from day to day, all of Mac OS X stays the same). If you've got 50 backups on the drive, and you have some file that has only had two versions during that time (e.g. the old content, before you changed the file a few weeks ago, and the new content, since then), then Time Machine will only have two copies of that file on disk (the old one, and the new one), for all of the backups -- not two per backup, just two copies total, with one linked into (roughly) half the backups and the other copy into the other "half". Thus Time Machine can be fantastically frugal with disk space. On the other hand, if you go through one day and, say, change the capitalization of artists names in all 20 GB of your music files in iTunes one day, well then that's going to chew up 20 GB on your backup drive on the next backup to save the new versions of all those files.
     
  11. hajime thread starter macrumors 68040

    Joined:
    Jul 23, 2007
    #11
    Thank you very much for the detailed explanations on TM. Could you please let me know why sometime we need to partition the drive if we use SuperDuper?
     
  12. Pax macrumors 6502a

    Joined:
    Dec 12, 2003
    #12
    Thanks for all this useful info. I really do need to get into TM. Need to buy a bigger disk though.
     
  13. Smoothie macrumors 6502

    Joined:
    Jun 23, 2007
    Location:
    California
    #13
    This is one of the most informative threads I've read. Thanks.
     
  14. hajime thread starter macrumors 68040

    Joined:
    Jul 23, 2007

Share This Page