Backup and Tape device answers

[Tim Jones]

Hi Folks,

I've been working with and designing backup tape hardware and software since 1984. If you have questions about backup tape technology, I can probably provide an answer or can at least point you to the best resource to find one. From old (ancient ?) QIC, DLT, or MLR/SLR to the latest DAT, LTO, AIT or SDLT, I've either worked with the design team, or with the physical devices. I currently deal with Mac OS X, all of the Unix variants, and Windows platforms.

While I work for a backup software company, I promise to keep my answers non-software product specific as I try and stay on top of the posts here.

Please don't hesitate to ask questions about backup tape hardware.

Tim

 

[Brian Timmons]

Welcome Tim,

Backup issues, particularly on OS X, are gaining much interest on the forum lately. I'm sure your expertise will be greatly appreciated.

Your company's software has also made life easier for many of us.

Thanks.

Brian Timmons
britim media

 

[michael zaletel]

Hello Tim:

It's been a long time since I've used any sort of tape-based media for backup and storage. I have one question for you. What is the current state-of-the-art in tape-based backup for backing up a 16TB RAID to tape once per week for off-site storage. Cost, speed, amount of media required? At $95 for a 1TB hard drive, isn't it cheaper and much faster to just build a backup RAID and store that offsite?

Please advise,

-shooter

-shooter

 

[albert rudnicki]

Hello Tim
I got a question for you.
I just received a Dell PowerVoult 110T DLT 3 external unit and I paired it with a ultra320 adaptec PCI-X
I have in total 3 workstations which are Mac and PC
My network is all 1 Gigabyte.
Since it's SCSI The lto will be running on the PC platform.

The question is : can I run my DLT on the network and with what software.
I usually run the backup at the beginning of the project (r3d files) and then at the end to backup the project and master files. I don't use it on regular everyday basis.


Thank you for your help and contribution to the forum !

Best
Albert

 

[Tim Jones]

Hi Shooter,

While disk may seem the "cheap" alternative to tape of late, disk comes with the expense of keeping it powered and what you're plugging them into. Plus, 16TB of RAID-5 backup (I'd use RAID-6) drives would require 21 drives (1,000,000,000,000 Bytes, not 1,099,511,627,776 Bytes per drive) plus 2 RAID parity drives AND a chassis capable of supporting 21+ drives - That's $1,995. With an LTO-4 drive on a similar system, you'd need ~$1,400 worth of tapes (20) per swap. And, that's assuming that you're not using server-grade disk drives which are MUCH more expensive than $95 each.

In fact, if you search for 16TB RAID-5 arrays. you'll uncover that to get RAID-5, you're only getting 13TB of usable space. For example, the Avid VideoRAID SR is around $22K and you're stuck at 13TB usable. However, an HP MSL 2024 LTO-4 library will set you back around $9K - $10K. Add the tapes and you're only at $13K max. Seems to me that tape's a win here if we're talking solely dollar costs.

[edit] Also, keep in mind that it can take 50 hours or so to prepare a 13TB RAID-5 array for use...

As for storage, when I take an LTO-4 tape and put it on the shelf, it can sit there for 5 years (or 50 years) and I can recover it. However, if I leave disks that had spun for some period on a shelf, for every 6 months they are left "unpowered", there's an increasing probability that they will suffer from a problem known as "Sticktion" that can cause the disk to not spin back up properly when it finally is repowered (this is the #1 cause of failure in LaCie drives, IIRC).

So, if you're not in need of long term storage (archival) or you can power the drive up weekly for a period of a few hours to reduce the possibility of Sticktion, and have taken the expense of storage and connectivity into account, disk may be a good answer for you. For me, my data's too important and I trust LTO and DAT for my long term retention needs.

To performance, we have users running LTO-4 media at consistently around 100MB/sec on Intel Mac's with Fibre-Channel or SAS connectivity. This would allow for the backup of 16 TB in 46 hours + tape changes. Or, if you can split the data into smaller chunks and run multiple drives (say 4TB to 4drives) with SAS or F-C you'll see an even more useful reduction.

Just my $.02

Tim

 

[Tim Jones]

can I run my DLT on the network and with what software.

Hi Albert,

Unfortunately, not as a device on the other systems - in other words, the Mac can't write to the tape drive on the PC. However, you could use client/server software that will use your PC as the server and treat the other systems as clients.

I'd look at Netbackup or such. Be prepared for the expense...

Tim

 

[albert rudnicki]

Just realize of two mistakes I made in my post. (not DLT but LTO-3

Is 1gig network fast enough to support lto-3 ?


Thanks Tim

 

[warwickt]

brumeister! wonderful to have you on the forum. your expertise and opinions are invaluable to we content creators!
w

 

[Tim Jones]

Just realize of two mistakes I made in my post. (not DLT but LTO-3

Is 1gig network fast enough to support lto-3 ?

Yes it is fast enough - with the right software (100MB/sec is attainable over that link). Glad you cleared it up about the DLT, we were all starting to wonder if you were still recording on Quad... :innocent:

Tim

 

[Tim Jones]

Thanks, Warwick. Glad I can share. After all, what good is knowledge if it's not shared?

Tim

 

[michael zaletel]

However, an HP MSL 2024 LTO-4 library will set you back around $9K - $10K. Add the tapes and you're only at $13K max. Seems to me that tape's a win here if we're talking solely dollar costs.

Ouch. $13k to backup 16TB to one set of tapes? There isn't a more affordable tape-based solution?

-shooter

 

[Chris Swinbanks]

Hi Tim.
I've got a library of Exabyte and Metrum tapes that......... no, no we won't go there... no Red files on them. :clown2:


Welcome Tim, look forward to reading your posts.

cheers
Chris

 

[Mark Phelan]

Tim, what's your opinion on the various flavors of LTO setups? LTO-3, LTO-3A, or LTO-4? LTO-3 is the cheapest, but has no directory like LTO-3A. LTO-3A has a good directory setup for retrieval, but is expensive. LTO-4 is the newest, but as you said, it's expensive.

 

[Dave Blackham]

Id go with LTO4, apparently LTO3A has compatibility issues.

 

[Jeff Kilgroe]

Welcome aboard, Tim!

Id go with LTO4, apparently LTO3A has compatibility issues.

LTO-3A is a compatibility issue. I personally recommend LTO-4. it's not quite as cheap as LTO-3, but it's faster, prices are falling like a rock, and is the current spec. Complete desktop SAS kits with some tapes and software can be had for under $2500. In the long run, if you are archiving 30TB or more over the next year, it's both cheaper and more reliable than HDD. It still boggles my mind how people trust shelf-destined backups to hard drives.

 

[michael zaletel]

It still boggles my mind how people trust shelf-destined backups to hard drives.

I've never ONE time, in 20 years had a hard-drive fail or go bad after sitting on the shelf for 5 or 10 years. Why is everyone so worried about that. Sure I hear stories but I also have had many situations where I could NOT retrieve data off tape backups because the unit failed and I couldn't find similar used unit or because operating system, drivers, indexes or some other incompatibility issue. Let's put it this way. I'd rather have 3 separate 1TB hard drive copies of my critical data (at $295 total cost) stored in three separate locations for 10 years than 1 tape backup that I may not be able to read 10 years from now.

Just my humble opinion.

-shooter

BTW, I sometimes have nightmares about Retrospect from the 90's.

See
http://reduser.net/forum/showthread.php?t=21284

 

[Jeff Kilgroe]

I've never ONE time, in 20 years had a hard-drive fail or go bad after sitting on the shelf for 5 or 10 years. Why is everyone so worried about that.

Because over the last 15+ years, I've purchased literally thousands of hard drives and had hundreds of them fail. Hard drives are not designed to sit. Platters are not designed to store data for extra long periods of time. Most HDD manufacturers recommend reformatting and reconditioning drives every couple years. If you have not had any troubles, then you are one of the rare and fortunate few. Seriously, how many hard drives have you purchased over the past 10 years? How many have you allowed to sit on a shelf with data for that period of time? Have you checked them all recently?

Sure I hear stories but I also have had many situations where I could NOT retrieve data off tape backups because the unit failed and I couldn't find similar used unit or because operating system, drivers, indexes or some other incompatibility issue.

All mechanical devices can fail. I've heard a lot of horror stories about tape over the years too. I've had drives eat tapes, I've had tapes go bad or have corrupt data and not explanation can be found. What I think is laughable is that people use compatibility or "inability" to locate a drive and software to read a tape at some point in the future as an excuse to not use tape. The same logic can be applied to any other format -- HDDs, optical media, etc.. All the good tape formats from 25 years ago are still readable today, many of them (like LTO) are readable in the newest LTO-4 drives.

Let's put it this way. I'd rather have 3 separate 1TB hard drive copies of my critical data (at $295 total cost) stored in three separate locations for 10 years than 1 tape backup that I may not be able to read 10 years from now.

I would prefer 3 HDDs to 1 tape copy any day myself. Why only use 1 tape? What kind of comparison is that? 1TB HDD's are $60 these days. 800GB LTO-4 tapes are $50. What's the point? The only argument against tape that I can see is that they are slower than HDDs and require an additional piece of hardware (drive). OK, fair enough. But what makes you think that in 15 years, you will just be able to connect that SATA drive to a common computer without additional hardware? 15 years ago, RLL, ESDI and 25-pin SCSI were all common drive standards. All would be a real PAIN IN THE ASS if you had to connect one today. Whereas I can put an LTO-1 or LTO-2 tape in my LTO-4 drive and just read it.

BTW, I sometimes have nightmares about Retrospect from the 90's.

I have nightmares about any version of Retrospect. I wish it upon all of my enemies.

 

[brainburst]

I agree about drives being the answer not tape. You can get decent drives cheaply today and it's not a huge deal to put them in a mirroring capable enclosure and spin them up once in a while. You have random access and an easy migration path. Corporate backup is moving away from tape in large numbers. Furthermore Compact Flash is getting cheap. We are getting to the point where we can just keep the cards and not overwrite them. San disk is working on write once media that will be great for archiving. You hardrives in a mirrored setup will last long enough to migrate to the better media.

 

[Jeff Kilgroe]

I agree that FLASH, especially write-once media will play a huge role in the future. I don't agree with the hard drive deal though. I'm not sure where you're coming up with "Corporate backup is moving away from tape in large number" statement. LTO-4 is selling like crazy and provides huge cost savings over drives for large archival volumes. When it comes to archiving petabytes and zetabytes of data, hard drives don't make sense at all.

SAN installations with online backups are predominantly hard drive based these days, and typically that is where "corporate backup" tech is migrating to, but this is not long-term archival.

I seriously do not see how hard drives can provide a realistic and simple solution to archiving anything in the PB range or larger. Look at the size, weight, additional storage containers, etc.. needed to store that many hard drives. Tapes occupy less physical space and weigh a lot less. Plenty of autoloader systems for tapes that can be had for a lot cheaper than hiring an IT school flunkee to swap drives into a RAID enclosure. Buying enclosures for every collection of drives becomes a huge burden in terms of cost and physical size as the archives grow.

I often wonder just how much data people on reduser are really intending to archive and for how long. I'm shelving duplicate copies of 1.2 to 1.8 TB every week or two. LTO-4 tape has already proven to be more cost effecting than hard drive over the past few months. I transferred all my HDD based REDCODE archives over to tape. Done. Out of the 500GB HDDs I was using to archive my RED footage and everything else I have done over the past year, I had one drive failure. Out of about 40 drives. And they were all less than 6 months old, bought and used just for backup. Had one bad tape, too... The drive failure was not an issue, as I had duplicate copies of all things. It was defective right from the factory, two separate tape drives would reject it with tape transport error.

I feel a little sorry for a few reduser members I know that have been going the HDD route, who had bought many of the Seagate 1TB units that are failing like crazy. Oh, well, shit happens.

 

[J Davis]

Larry Jordan, esteemed final cut guru has this to say about hard drives.

Magnetic signals recorded on a hard disk are designed to be refreshed periodically. If your hard disks stay on, this happens automatically. However, if you store your projects to a removable hard drive, then store that hard drive on a shelf, unattached to a computer, those magnetic signals will fade over time... essentially, evaporating.

Link to his article
Hard Disk Warning! by Larry Jordan
http://www.larryjordan.biz/articles/lj_hard_disk_warning.html


EDIT
I know cost of LTO is an issue for a lot of us so I started a thread called
HP DAT as a backup alternative to LTO
http://reduser.net/forum/showthread.php?t=24988