System Image vs 'normal' backup of system drive

[Simon]

As I've said before, all my data, program files etc., are automatically backed up hourly in the background to an NAS, so I really can just forget that side of things. And TBH, I really can't see any advantage to doing it any other way.

What type of backups do you do?  Are they all incremental following an initial full backup, and if so, how many incremental backups do you do per one full backup?

[colirv]

None of the above! I splashed out $60 on two of these for mine and my wife's PCs. There may be cheaper or free alternatives around. Genie simply checks every so often (you can set the period) for any files that have been updated, and copies them to the drive you specify. I use an NAS, and I can go to it any time I want and find the latest version of every file, in its original format. I've got it set to save earlier versions of files as well, going back a week.

The very first backup took about a day in all (at least), as all the files had to be sent over the wifi. Once that was done, however, the continuous backup is not noticeable - a few seconds in the background every hour depending on how many files you've changed. If you ignore the option to keep multiple copies the effect is as though you went to Windows Explorer every hour and told it to copy your whole drive to another disk, ignoring files that haven't been changed - except that would be somewhat of a chore to do yourself (even if you used programs like Synctoy or FreeFileSync), whereas Genie it does it for you!

[Simon]

OK, thanks.  I'll take a look.   

[armadillo]

The bit I've bolded is the bit which I'm not getting, but can I use this to combine all the incremental backups into the initial full backup, and therefore reduce the number of incremental backups building up?

Acronis terrible help file strikes again! It means you can combine all the incrementals with the full backup and then delete the incrementals. However, as you have found, it is useless as it takes longer than making a new full backup.

[Simon]

Yes indeed, I gave that up as a bad idea!

[armadillo]

It was just all the incremental folders piling up which is worrying me a bit, but I don't want to go to the 5:1 thing (where it does a complete backup for every 5 incrementals), as I don't think that's necessary.  I know I can adjust that to whatever I want, so that may be the best option, but I was trying to tidy things up a bit, and get everything so I could 'set and forget' it.

The best way is to do a full backup followed by differentials, not incrementals. Differentials do not form a chain so all you need for recovery is the full backup plus any one of the differentials. You can delete any number of differentials as you wish, without compromising the integrity of the backup. Differential backup is fast and, unless you are updating a large proportion of your data, the backups are tolerably small too.

[colirv]

I can think of two problems with differential backups.

a) They take up more space than incremental backups, which could be important, particularly if you're struggling to store two full backups (which is preferable, otherwise you're without any backup while you're doing the new full backup).

b) You will only have one version of each file you've been working on. Previous versions, which might exist in incremental backups, are over-written by each differential backup, which uses the current version. This, incidentally, is one reason why I use the different, automatic method of backing up files that I described above. I can go back an hour, or up to a week, if I make (as I have done in the past) a mess of something!

[armadillo]

I can think of two problems with differential backups.

a) They take up more space than incremental backups, which could be important, particularly if you're struggling to store two full backups (which is preferable, otherwise you're without any backup while you're doing the new full backup).

They do take up more space than incrementals but how big they are depends on how much data has been altered since the last full backup. So it might take many differentials before the same space is taken up as a further full backup.

b) You will only have one version of each file you've been working on. Previous versions, which might exist in incremental backups, are over-written by each differential backup, which uses the current version.

Previous versions are not overwritten by any differential backup. When you make a new differential backup, it does not alter the full backup nor any of the previous differential backups. You can always go back to a previous version (provided you have not deleted it).

For example, suppose you make a full backup on day 0 and then differentials on consecutive days, d1, d2, ..., d10. If you want to restore the data as at day 5, you just restore from the full backup together with differential d5. (Maybe I misunderstood your point. Forgive me if I did).

You have just as much flexibility as with incrementals. But to do that with incrementals, you would need the full backup plus the incrementals from each of the five days and you would be stuffed if any of them is corrupt.

[Simon]

The best way is to do a full backup followed by differentials, not incrementals. Differentials do not form a chain so all you need for recovery is the full backup plus any one of the differentials. You can delete any number of differentials as you wish, without compromising the integrity of the backup. Differential backup is fast and, unless you are updating a large proportion of your data, the backups are tolerably small too.

I don't quite follow when you say differentials don't form a chain.  Do you not still get "BackupA", followed by "BackupA1", BackupA2", etc?  I get that it's saving changes since the original backup, rather than since the last incremental, so does each differential overwrite the previous one?

[armadillo]

The differentials are completely independent of each other and do not form a chain. No differential overwrites any other. Backup A remains unaltered and valid. Each differential remains unaltered after it is created

At the time you create BackupA8, the total state of all your data is given by BackupA and BackupA8. BackupA contains all your data in its original state and BackupA8 contains those parts of your data that have changed since then.

BackupA1 to BackupA7 are no longer needed unless you want to recover to the time when one of them was created.

Imagine that your data consists of 100 document files, say f1, f2, ...., f100.
BackupA contains a (compressed) copy of all 100 of those documents.

On day 1, you work on files f20 and f50, say.
So BackupA1 contains not all 100 files but just the updated f20 and f50.

On day 2, you work on files f30 and f35, say.
So BackupA2 contains not all 100 files but just the updated f20, f50, f30 and f35.

On day 3, you work on f27 and f22.
So BackupA3 contains not all 100 files but just the updated f20, f50, f30, f35, f27 and f22

To recover your data as at day 3, the software copies all 100 of your files from BackupA and then updates them with the 6 files that are held in BackupA3.

Differential disk backups work in the same way except that they record altered disk sectors rather than altered files.

[Simon]

Right, I'm with you.  Thanks!   

[colirv]

Maybe I misunderstood your point. Forgive me if I did.

No, you didn't. I was wrong. I still think, however, that although imaging has a place in terms of recovering from a disaster, and one can backup just the system to do that, using imaging simply to keep copies of files seems unnecessarily clumsy. Perhaps it's me that's misunderstanding.

[armadillo]

No, you didn't. I was wrong.

That is generous of you.

I still think, however, that although imaging has a place in terms of recovering from a disaster, and one can backup just the system to do that, using imaging simply to keep copies of files seems unnecessarily clumsy. Perhaps it's me that's misunderstanding.

I agree that imaging is a great way to backup the system but a clumsy way to backup files. However, you may be misunderstanding what the imaging software, such as Acronis True Image, EaseUS Todo and a host of others, is doing with file backups. When it makes file backups, it does not actually create images at all.

When the software creates system image backups, it makes backups (full, incremental or differential, as you choose) of the disk sectors of the whole disk. Those backups are system image backups. They are the most reliable method of recovering a bootable system. If the computer is designed from the outset so as to keep the OS on a dedicated disk, this is very efficient in terms of time and storage.

For backing up data files, it is not usually best to create disk images of the disks which store the data files. But that is not what any of the imaging softwares do anyway. What Acronis (and EaseUS and all the others) do for file backups is to create backups merely of the files, not of the disk sectors. The same principle of full, incremental or differential is applied to these file backups.

When you specify file backups to Acronis True Image, you select the folders or files you want backed up and it then creates full, incremental or differential backups just of those specified files. Those backups are therefore small and very fast. They are not "images" in that they do not record any structure at all, just the data.

Acronis (and the others) started out by providing just partition and disk imaging in their software. Then they realised there was a market for the simpler file backup for data and they added file backup features to their software.

So, every backup is:

(Disk image level OR partition image level OR file level)
AND
(Full OR incremental OR differential)

The software can create any of those.

[colirv]

Thanks for that!

[Simon]

It's more complex than you might think, isn't it!  I've done fiddling with Acronis now, so I'll leave it for a month and see what it does.  I've set up some auto-deletes, so it will be interesting to see if they work according to plan.