![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
What IS a backup
- Thread starter zaniix
- Start date
BunnyKillBot
n00b
- Joined
- Mar 25, 2012
- Messages
- 33
I have seen that floating around here too and I think they are confusing backup vs archive. It can pretty much be summed up as online(live), nearline(backup) , offline, (archive) differentiated by performance vs availability. For example in a business context, your local drive for live data, a NAS for backup and LTO for long term storage.
A backup is something that decreases single point of risk.
Having your backup on the same system as your primary data, is normally a major issue, due to common faults of power supplies, electrical, disk failures, admin mistakes.
Having a seperate system, can limit you to common building (electrical systems, fires, earthquakes, ...) and would take multible admin mistakes to have problems. These can be solved with offline backups.
The problems above can be solved by going with offsite backups, be them online or offline.
Having your backup on the same system as your primary data, is normally a major issue, due to common faults of power supplies, electrical, disk failures, admin mistakes.
Having a seperate system, can limit you to common building (electrical systems, fires, earthquakes, ...) and would take multible admin mistakes to have problems. These can be solved with offline backups.
The problems above can be solved by going with offsite backups, be them online or offline.
There is a 3-2-1 rule for backups that says:
Have three copies of your data on two different types of media (tape, hard drive) and have one copy offsite.
To me, a backup is a second copy of your data that is in a physically different place then the source copy.
Have three copies of your data on two different types of media (tape, hard drive) and have one copy offsite.
To me, a backup is a second copy of your data that is in a physically different place then the source copy.
Technologies aren't backup. That is, if you use RAID 1, it's not backup. It will minimize the chances that a single drive failure will be disruptive.
For backup for home, I've done the following.
1. Determine what files I really want to protect and provide both local copies and remote copies both in the cloud (more than one service) and given to family members for offsite backup. This is primarily photos and other electronic media.
2. Use a Windows Home Server to backup the entire systems. Again, this is more to minimize disruption so that I can replace a failed drive if needed or undo a bad software update. I rarely have needed to do this.
3. Backup other files to a NAS and a backup of the WHS.
Whatever method or methods you use for backup, you have to make sure you actually use it consistently.
For backup for home, I've done the following.
1. Determine what files I really want to protect and provide both local copies and remote copies both in the cloud (more than one service) and given to family members for offsite backup. This is primarily photos and other electronic media.
2. Use a Windows Home Server to backup the entire systems. Again, this is more to minimize disruption so that I can replace a failed drive if needed or undo a bad software update. I rarely have needed to do this.
3. Backup other files to a NAS and a backup of the WHS.
Whatever method or methods you use for backup, you have to make sure you actually use it consistently.
Traditionally, backup is pure disaster backup.
If something happens you have an external backup of last data state ex yesterday and in
best case a second with a former state.
A raid like a mirror does not help as on an unwanted file delete or a file encrypt
by a trojan or virus this happens on both disks of a mirror.
For such problems a disaster backup does not help quite often as you detect the problem too late.
While you always need a disaster backup (an external copy that survive a disaster like a fire),
a modern backup concept must offer more. A disaster backup is not up to date enough or does
not offer a previous version on a former state. So you need a
- undo a delete or unwanted modification based on versioning (down to a short interval versioning ex every hour)
- prevent versions from modification (they must be readonly by concept)
- offer versions over a long time like days, weeks, months or more
- offer a mechanism to verify and repair data on disk based on checksums, either on access or on demand
So a modern backup concept needs a storage that you really can trust with versioning and auto repair. It needs to
be crash resistent and must offer redundancy, not only for uptime but mainly for auto-repair.
This is where you need the 3rd gen filesystems with copyonwrite + data/metadate checksums +
a scrubbing (online repair based on checksums) and raid (redundancy for repair) like btrfs, ReFS or ZFS.
If you combine this with filesystem replication with a versioning to one or two external backups systems,
you are protected against nearly any dataloss option with the option to go back to "last month, first friday morning"
on your primary storage (ex with Windows previous versions)
If something happens you have an external backup of last data state ex yesterday and in
best case a second with a former state.
A raid like a mirror does not help as on an unwanted file delete or a file encrypt
by a trojan or virus this happens on both disks of a mirror.
For such problems a disaster backup does not help quite often as you detect the problem too late.
While you always need a disaster backup (an external copy that survive a disaster like a fire),
a modern backup concept must offer more. A disaster backup is not up to date enough or does
not offer a previous version on a former state. So you need a
- undo a delete or unwanted modification based on versioning (down to a short interval versioning ex every hour)
- prevent versions from modification (they must be readonly by concept)
- offer versions over a long time like days, weeks, months or more
- offer a mechanism to verify and repair data on disk based on checksums, either on access or on demand
So a modern backup concept needs a storage that you really can trust with versioning and auto repair. It needs to
be crash resistent and must offer redundancy, not only for uptime but mainly for auto-repair.
This is where you need the 3rd gen filesystems with copyonwrite + data/metadate checksums +
a scrubbing (online repair based on checksums) and raid (redundancy for repair) like btrfs, ReFS or ZFS.
If you combine this with filesystem replication with a versioning to one or two external backups systems,
you are protected against nearly any dataloss option with the option to go back to "last month, first friday morning"
on your primary storage (ex with Windows previous versions)
When storing archival backups in deposit boxes, closets, etc., keep in mind that both SSDs and HDDs should be powered up every six months to prevent data loss. You old-school HDD proponents who think HDDs are invulnerable to power-off time, be aware that they don't make HDDs like they used to.
evilsofa, can you expand upon what you are referring to? If a hard drive is in a safety deposit box for two years, and hasn't been powered up once for the whole two years, are you saying it won't work? Or data could be lost, missing, or corrupt? How does this happen?
Yes there can be data loss becasue magnetic states change and without the drive on to correct any issues there can be loss.
This is pretty much the major issue with any backup, so many ways for the data to be lost.
westrock2000
[H]F Junkie
- Joined
- Jun 3, 2005
- Messages
- 9,434
What happens is that people think hardware redundancy (RAID) is backup. RAID is only to protect you from a certain amount of mechanical failure. However RAID will not protect you from accidently hitting delete key or from more drives than allowed failing.
2 disks failing on a RAID5 setup means complete data loss. The chances of 2 disks failing are much less then 1 disk failing though.
So generally this debate springs up when people are looking at RAID and unknowingly calling it backup, which agitates some people.
So for instance I have RAID with 2 drives serving as redundancy on my movie server. But I do not have an actual backup in place outside of re-ripping the discs themselves. So I am protected, generally, from hardware error. But if I accidently delete something, it's gone forever. This is what people are trying to get across.
A backup protects your data from being lost. RAID protects you from downtime.
2 disks failing on a RAID5 setup means complete data loss. The chances of 2 disks failing are much less then 1 disk failing though.
So generally this debate springs up when people are looking at RAID and unknowingly calling it backup, which agitates some people.
So for instance I have RAID with 2 drives serving as redundancy on my movie server. But I do not have an actual backup in place outside of re-ripping the discs themselves. So I am protected, generally, from hardware error. But if I accidently delete something, it's gone forever. This is what people are trying to get across.
A backup protects your data from being lost. RAID protects you from downtime.
westrock2000
[H]F Junkie
- Joined
- Jun 3, 2005
- Messages
- 9,434
You can also look at BD-XL bluray discs. You can get 50GB pretty easily at this point. It's more expensive, but you do not have the power on issue. And you can even look at "M-Disc" blurays which are supposed to last a REAAAALLLLYYY long time.
There's some good answers in this Q&A from 2011. They suggest refreshing data and making sure you can still read it every 2 to 5 years for three main reasons: magnetic field breakdown, environmental conditions and mechanical failure.
I suggest being more conservative now than they were even 5 years ago because areal density on HDD platters has gone up significantly and continues to go up (see: mentions of the Magnetic Recording Trilemna in the wikipedia entry for HAMR), and also because there remains the boogeyman of strong magnetic events such as the solar storm of 1859, which hopefully will not happen in our lifetimes but are impossible to predict (we had a near hit in 2012). I expect that in such events, or the much more likely lesser events, the amount of magnetic field breakdown could make a difference in the amount of recoverable data.
Nenu
[H]ardened
- Joined
- Apr 28, 2007
- Messages
- 20,315
A backup is having your data you dont want to lose stored somewhere it can be recovered from.
I have to say, this is a stupid question.
If you meant a different question, its worth asking that.
I have to say, this is a stupid question.
If you meant a different question, its worth asking that.
A backup is "offline", meaning it is not actively in use.
Like if you have a mechanical pencil. You can do things to make one pencil more reliable, but in the end sometimes that pencil can break from use. In those cases it is nice to have another pencil in your backpack still in its package ready to go
Like if you have a mechanical pencil. You can do things to make one pencil more reliable, but in the end sometimes that pencil can break from use. In those cases it is nice to have another pencil in your backpack still in its package ready to go
A backup can be many things, but most people are usually saying that something isn't a "good" backup.
A good backup needs to take measures to protect against many things.
It needs to protect against hardware failure (hdd failure, ram, etc), software failure (data overwrite, corruption) environmental disasters (fire, theft), even user error (accidental modification, deletion).
So a backup needs to be a separate copy of the data on another system, the copy needs to be in a physically different location, and it needs to employ a form of version control.
There can be more to it than this too.
A good backup needs to take measures to protect against many things.
It needs to protect against hardware failure (hdd failure, ram, etc), software failure (data overwrite, corruption) environmental disasters (fire, theft), even user error (accidental modification, deletion).
So a backup needs to be a separate copy of the data on another system, the copy needs to be in a physically different location, and it needs to employ a form of version control.
There can be more to it than this too.