- Tape is essentially sequential. A file at the end of a high-capacity tape takes a while to access

- Tape degrades over time far more than disk. A tape backup done on Monday may not be readable on Tuesday - and you won't know until you try it

- Tape is probably more expensive than disk

- Tape is less accessible than disk

- Disk images can often be mounted as a separate drive

I can think of a unique characteristics of tape (while without tape library), that is it can be kept offline.

Keeping a copy of data offline may help in case there is need to investigate some sort of fraud by IT insider who may intends to wipe off any traces with their knowledge.

With an offline copy, there is a big obstacle for any  techo guru, as they cannot do anything unless they get access to that copy, which probably needs to go through series of internal control.

@Paul Thornett said:

- A tape backup done on Monday may not be readable on Tuesday

!!!!

I don't think I'm sticking my neck out here by saying that Paul might be overegging the instability of tapes here.  True if someone came up with a biodegradable tape that can be buried in landfill and break down completely to worm food in 6 months then his statement may be true in general.  

The same statement could be applied to disks though - who hasn't experienced a flash drive or a hard disk corrupting a file occasionally - does this mean we should be worried that files may not last more 20 seconds on disks!  

Both mine and Pauls statements are scaremongering people into choosing a particular solution we prefer but we have to understand why these options are chosen and 99% of those are 'MONEY'

The major benefit we have with tapes is that they are easy to take images of all of our servers and store offsite in a safe cheaply.  We only have one server room and for disaster recovery purposes we need backups away from this server room.

Large companies with multiple server rooms may have a completely opposite opinion to storing data on tapes to us as they can mirror their backup data across different sites.

Rarely do decisions come down to 'What makes our lives in IT easy?', they usually come down on the side of 'What is cheapest to run?'

I don't use tapes for daily backups anymore. I use an online offsite backup service.

Backups happen on schedule, multiple generations are saved, quick restore, no one forgets to take the tape off site each day, etc.

There is a cost to this but when I did an analysis of the costs of tapes, drives, labor, etc over 5 years, the cost of online offsite data storage was not that much more. And now it is a hands off operation.

I have found a WIDE range of costs for online offsite data storage, so shop around. I pay around $250 a month to store approx 90GB of data.

Each media has its own specific area of application. By themselves, neither tape or disk provide a total backup, recovery, retention and audit solution. So far the discussion has omitted other solutions (like SAN-based offerings from Veritas and EMC); however by themselves those are not solutions and just create additional failure points.

Everyone who reads this has been burned by their backup "solution": the backup was bad or missing. Mitigation is the goal.

Steve's original point is really about mitigation and that the various backup methods must be reviewed in concert to create a requirements-based solution. If you will never have disk or tape failure, don't worry about redundant backups. If your data center will never burn down or be compromised, don't use offsite backups. However, the  redundancy of disk AND tape will provide a simple, cost effective and robust solution. I myself like to backup the database to disk and then backup disk to tape (or SAN or ...?).

Let's not spend $1 million to create a pen that works in outer space when a 2 cent pencil works fine. :-)

Steve, thanks for the musings!

Being responsible for 45 School's Databases I sleep better at night using SonicWALL CDP applainces. No Fuss No Muss. It is a disk based solution with encrypted offsite backup availaable (this is not like those IOmege drives I tried that and they failed). The CDP devices back up SQL Server Databases, Exchange server, Active Directory, PST files, as well as clients Desktops with no daily intervention and can be enabled to aslo back up off site. Restoring to several previous versions is so simple.  

I use off site disk storage for backups.  I bought a server with 2TB of storage and it cost less then the Tape backup device.  

Also at a previous employer we had a tape drive get its write head out of alignment and our backup location could not read the tapes.  We Got it fixed because we checked the tapes on a regular basis, but the tapes were worthless the moment we took it out of the drive.

I am using a disk to disk with offsite backup. The advantage is, I get high speed backups from all of my servers to a central disk array over local fiber or cat5. Then I can transmit those to a remote offsite backup at any time I schedule. I can also throttle the bandwidth, if the client is using the line 24-7. In terms of cost, I stopped buying pre-built arrays and started building my own, because I couldn't find anybody that could build a backup array that was any better than I could build at even twice the cost. Using BSD on CF cards I can pre-build configurations for a client and have them stored and easily deployed much like I would on CISCO equipment. SATA drives in RAID 5,6 or 10 allow me to have multi-terabyte backup systems for less 1K and I don't have to worry about where the tape went when I travel onsite. I think tape still has a use in specific situations, but it is quickly being replaced by disk.

We created our own backup solution, using unc paths on the domain.  Also, we use drive systems similar to RDX @ half the cost.  We swap it out daily, do disk scans and file verification Every Friday. The system itself handles the backup job in about 2 1/2 hours.

Having spent a better part of my career working with 6250 and earlier tape backups -spanning 26 reels at one time for some hierarchical DMSII  backups - you can only image how I feel about tapes.

They're not environmentally favorable and I for one would be happy if they would go away for good.

Now if you want to talk card decks :)

VTL for the win! Archiving to tape with a policy means no mess no fuss automagic.

I don't think multiple hard drives spinning away all day and all night are particularly environmentally friendly either. At the moment a lot of companies are moving away from tapes to disks, but the cost savings are not clear.

If the point-of-failure of tapes was forgetting to change the tape, then get a robot. If you have the bandwidth to move data onto remote hard drives, then you can equally move the data to remote tape drives. Both tapes and disks fail, given time. Disks can be arranged in nice arrays which prevent data loss when a single drive fails. Data can also be copied to more than one tape (wow, there's a thought).

For me, a massive advantage of tape is when you want to increase the size of your archive; just buy a few new tapes (unless your robot has run out of space). For these large disk arrays, whilst the cost of a disk is cheap, the enclosures for the disks are not, and adding extra capacity is nowhere near as simple.

Let's face it, no backup system is 100% full proof, but there are ways to greatly reduce the risk of data loss.  The best solution for us is a combination of tape/disk backups.  Critical daily data is backed up to tape every night and goes offsite.  Regular daily backup is done to disk.  Monthly archival is done to tape for long term storage.  This scenario gives us a flexible yet secure policy with little fear of data loss.

One of the things that floors me when people turn to Disk only instead of Tape is that they forget that Replication IS NOT A BACKUP!  I once got into a somewhat heated debate with a customer's CIO about this as he felt they were covered because they had their centeras setup in a mirrored configuration between their DR and PROD.  What more of a backup do we need.  He still stood by this even after they replicated deletes that were not supposed to be run in the first place.

A true to disk (with some sort of offsite\retention)  can be a very viable solution.  Just remember.  Plan for the possibilities.