Server Backups

[USMCGrunt]

Not really sure where this question goes, so I'll start here....

I have a new organization that is thinking about putting in a server and a 12TB NAS. They want a full backup of all data conducted daily and the backup to be able to be moved into a fireproof/waterproof safe. I've done this kind of thing with smaller amounts of data using CRU DataPort products but 12TBs plus a couple hundred GBs in server backups is kind of beyond the abilities of what DataPort offers without going crazy on 3 or 4 HDDs being changed out daily.

How do large corporations manage something like this? What about Cloud based backup systems, how do massive sized backups like this manage to move across the cloud? Local ISPs have uploads limited to about 5-7mbps...it would take weeks for an initial backup to complete...not to mention the GBs of data change on daily incrementals. I'm obviously ignorant to the solution, please educate me.

 

[schizrade]

Not really sure where this question goes, so I'll start here....

I have a new organization that is thinking about putting in a server and a 12TB NAS. They want a full backup of all data conducted daily and the backup to be able to be moved into a fireproof/waterproof safe. I've done this kind of thing with smaller amounts of data using CRU DataPort products but 12TBs plus a couple hundred GBs in server backups is kind of beyond the abilities of what DataPort offers without going crazy on 3 or 4 HDDs being changed out daily.

How do large corporations manage something like this? What about Cloud based backup systems, how do massive sized backups like this manage to move across the cloud? Local ISPs have uploads limited to about 5-7mbps...it would take weeks for an initial backup to complete...not to mention the GBs of data change on daily incrementals. I'm obviously ignorant to the solution, please educate me.

I am using Dell AppAssure with their DL4000 appliance. A single base image (for us takes about 1.5 days for 25 TB), then incrementals for eternity. It creates full backup sets in the database, without having to actually pull every TB weekly. EMC and others offer the same type of solution. I have incrementals running every 10 minutes from m-f 7-7 and hourly all other times. Dedupes to the DB/Storage unit. I replicate the changed files to a remote site nightly that hold an identical copy of the DB. I also have a LTO6 tape library that rolls a full snapshot every 2 weeks to send away out of area.

Offsite replication is key to quick disaster recovery, and tape is for after the zombie apocalypse.

One thing I always bring up to people wanting more storage, is that you have to back all of that up. You want 12TB of SAN storage? You better have a way to back it up. And when you start talking 12TB on a san with 1GbE ports, you better stop talking about weekly fulls, and start talking about constant incrementals and artificial full points for retention.

 

[bigdogchris]

You'd be better off having the NAS in a different building than the servers are in. If a plane hits the building it's not going to matter if the disk are in a safe or not.

There's zero need to swap disk as long as:

1. NAS is secure
2. There's enough room for multiple backups
3. It's in at least RAID 5 mode to allow a disk failure

There is also no need for full backups every day. That means a significantly larger backup window and a shorter retention. Do weekly fulls (over the weekend) and daily incrementals.

 

[USMCGrunt]

You'd be better off having the NAS in a different building than the servers are in. If a plane hits the building it's not going to matter if the disk are in a safe or not.

There's zero need to swap disk as long as:

1. NAS is secure
2. There's enough room for multiple backups
3. It's in at least RAID 5 mode to allow a disk failure

There is also no need for full backups every day. That means a significantly larger backup window and a shorter retention. Do weekly fulls (over the weekend) and daily incrementals.

This is for a small theater company of only about a dozen employees on staff but half of them do 2D/3D/AutoCAD work which is why there is a need for so much space but also that there is only one building with no funds or ability to run a network into another building.

I've also been under the impression from reading that RAID 5 is no longer recommended to use on these larger capacity disks. While it's been awhile since I read it, the idea is that if you do have a disk failure, you're very likely to encounter an unrecoverable read error in one of the other drives during a rebuild resulting in a total loss of the array. RAID 6 or 10 is now being recommended for large arrays.

There's a need to swap disks in the event of a catastrophic disaster such as fire, flood, airplane crashing through the roof, etc... Without a second premise to put another NAS for secondary backups that leaves us with what, putting another NAS in a different part of the building and hoping the catastrophe doesn't effect both places? The internet connection is also terribly slow on the uplink, somewhere in the neighborhood of 5Mbps, so full/incremental upload to the cloud wouldn't exactly be fresh data. In the event of a catastrophe, they want to have lost, at most, 24 hours worth of data. We are also trying to do this in-house, not rely on a third party offsite backup vendor if possible. So you see, we need a way of relocating a previous day/week backups to an alternate location in the building that can be fireproof/waterproof to minimize catastrophic impact....though I'm not sure a fireproof/waterproof safe will be....safe from an airplane crash.

 

[Mackintire]

I am using Dell AppAssure with their DL4000 appliance. A single base image (for us takes about 1.5 days for 25 TB), then incrementals for eternity. It creates full backup sets in the database, without having to actually pull every TB weekly. EMC and others offer the same type of solution. I have incrementals running every 10 minutes from m-f 7-7 and hourly all other times. Dedupes to the DB/Storage unit. I replicate the changed files to a remote site nightly that hold an identical copy of the DB. I also have a LTO6 tape library that rolls a full snapshot every 2 weeks to send away out of area.

Offsite replication is key to quick disaster recovery, and tape is for after the zombie apocalypse.

One thing I always bring up to people wanting more storage, is that you have to back all of that up. You want 12TB of SAN storage? You better have a way to back it up. And when you start talking 12TB on a san with 1GbE ports, you better stop talking about weekly fulls, and start talking about constant incrementals and artificial full points for retention.

+1 what he said.

 

[Deleted member 126051]

If they're going to do swappable backups. Tell them to just get a safety deposit box at their bank. Then rotate backups into the safety deposit box.