Advice on setting up an LTO system for backups

Hi all. I'm looking for some advice!

My boss would like me to look into setting up a 'proper' backups system. We're a small video production company with just over 1TB to be backed up. However, not much of this changes day-to-day. Most of the media that changes daily is on our Avid Unity server which is a big mirrored RAID (designed to protect data if single drives etc die.. or something!). All that media is all on source tapes anyway so we're not worried about that. What we want to backup is other media/graphics files, and general admin files, all on a variety of shared computers/drives in the office. I'm only guessing wildly, but probably out of that 1TB of data, it could fluctuate from ~10GB to 250GB changing in any one day. It all depends on what jobs we're doing, and what's going on.

At the moment all backups are manual; we have a server upstairs where I copy over all the essentials like admin stuff, emails, some media/graphics etc.. I also burn some stuff to disc. All this is what we want to automate, preferably to LTO. Our boss lives in Jersey so i'm thinking that'd be an ideal place to take away the tapes!

I've looked into LTOs and I understand that for media companies (and others) they are pretty much the standard. What I can't make a decision on is what kinda of drive to get, as I can't really figure out how much we'll be backing up every day/week! What seems to be recommended is a full backup over the weekend, then incremental backups every evening. This sounds great, but if we're backing up over 1TB every weekend, we're going to need an autoleader/library LTO drive to use multiple tapes aren't we? It wont fit on one, and I sure as hell am not coming in to swap tapes!

As far as software goes, I used to work in a large post production facility where they had a large number of LTO drive/autoloaders working with Backup Exec. It seems to crop up on here quite a lot, is this generally the one to go for?

Also, a real silly question; How do the LTO drives connect up to our network? I know our Unity is a fibre optic cable to each editing machine, so I don't know if that's possible? I presume it'd be good to do it that way, but failing that is it just regular ethernet/LAN?

I'd appreciate any advice, I'm kinda winging it here!

Couple of questions.

Is the data you are backing up centralised? If not, this complicates things a little. From both a planning and timescale of backup perspective. Trying to get a grasp from your post and it seems that most of what you want backed up is residing on desktops all over the shot and copied manually. I would highly suggest looking at centralising this data if it is viable.

LTO4 will net you 800GB of uncompressed backup with up to 1.6TB on a 2:1 ratio, you will not get 2:1 but the level of compression you achieve will largely depend on the kind of data you are backing up.

I am not a fan of incremental backups however I live by a rule of thumb - Spec tapes for native capacity with some headroom, any compression you achieve is a bonus and merely increases the length of time you can have it in service. Compression is not guaranteed. Others might dis-agree, but without having hardware to test your mileage with, it could cost you a lot of money on a solution that will not do the business.

I'm in Jersey myself, so hi2ur boss but keeping tapes that off-site kind of indicates to me your backup is merely as a fall back as opposed to a critical disaster/fast recovery and I am assuming getting them back over with a delay is acceptable?

2 of the main products out there at the moment are CA ArcServe and Symantec Backup Exec. Both are pretty much the same feature and cost wise with an array of agents. BackupExec has some Exchange granular technology which is supposed to be the bomb but I have yet to see it work as advertised and may well be surplus to your requirements anyway. Total cost of implementation with largely depend on what Agents you need to run on-top of the base product, again - centralising your data here will simplfy and lower the cost of this exercise. While either product will backup from network shares it starts getting messy when you do not use the agents.

1TB is a fair amount, are we talking lots of small files or lots of large ones, mixed bag perhaps? Tapes are great for long sequential writes but throughput will go through the floor on lots of small writes. 1TB of the latter will probably want to see you do some kind of disk staging or image based backups.

Go for LTO4 regardless of whether you chance it with compression, autoload multiple tapes or try incremental. You will still need a good Full backup to go the incremental route so in my mind you should avoid it. The price premium of LTO4 over 3 or even 2 is not that great.

You need to think about a decent HBA to run the drive off and whether you are going internal or external if you chose a single tape solution or what size autoloader you will go for. You really do need to plan and organise your media and backup strategy well if you go the Autoloader route. I've had to keep on-top of 24 tape autoloaders with seriliased labeling schemes and quite frankly it can be a damn nightmare! Gets confusing very fast if you do not keep track of it.

I think your first port of call should be identifying whether all the estimated 1TB of data needs to be backed up, or whether you can get away with a month end/quarterly/yearly archive of totally unchanged data that you just keep to one side. If it never changes, putting it onto tape every day is a bit pointless AS LONG as you have some kind of an archive of it. Tapes are resilient if stored properly.

As for connecting your tape drive, you will need a nominated box to run the backups from. Options are pretty flexible. SAS, SCSI, internal, external etc.

You could consider a compromise between pricing and automation and go for a single LTO4 solution with a full 2 tape backup on a Friday (if the compression does not cut it) with a manual intervention for a tape change and then incremental Mon > Thursday. You could start getting clever and doing a "safe" selection to one tape on friday, and a second selection to disk for later backup during working hours on Monday where you can have a physical presence. You could even stage the whole thing to disk and dump it onto 2 tapes during the day. This will have a performance impact on your backup server however and demand you have the free space on disks to place the backup.

If your data is not centralised then I would not even consider incremental backups.

Perosnally I would try and do a full backup each night. You can get 1Tb onto an LTO4, I am squeezzing 1.17Tb onto my tapes each night, they are FC connected to by cluster server, and that is in turn 2Gb FC connected to my SAN.

Backup takes 7hrs, and then a further 3 for the tape verify. Its all down to your infrastructure really.

Plus on a side note, if I had to go full & other type, perosnally Full & Differential is my preferred route. why waste a LTO4 with a few Gb of data, might as well get the tapes used, and then you have more chance of retsotring any files.

I'd first concentrate on sorting out the storage before trying to back it up.

Get a proper file server in place and redirect My Documents and prevent storage on desktop - that way everything should be stored on the file server.That will make backing it up a lot easier - you could then directly attach an LTO drive to the server.

Hi guys. Thanks very much for the info. Please bear with me as i'm no sys admin or IT tech! To answer the main point, no our files are not all centralised. We have one desktop with over 1TB of space which we use as our Media server. This is essentially the 'server' for our creative workings. Graphics, renders, music, DVD authoring.. pretty much anything that isn't stored on our Avid Unity. This is 2 drives which are shared across the office. As for filesizes it's a total mixed bag. There are some videos >1GB but it's not common. The rest is a hell of a lot of PSDs, DVD Authoring files, other images, etc..

On top of that we have two other desktops with 'admin/office' shares which are for our office team (i.e. non-creatives), with basically all admin files.. Word docs, spreadsheets... all that junk. So in total that is 4 different drives across 3 different desktops. Having said that, anything in My Documents on any of the workstations is probably not worth worrying about as it'd be personal.

Sin_Chase you're right in that the backup is at the moment just a fallback. It's more in case the whole building sets on fire or something! Having said that, if one authoring file is a week old but we've been working on that all week, that week-old file is next to useless! That's why i was thinking about incremental backups to catch anything that changes. Having said that, as it stands we only do a backup of our Avid projects (i.e. the edit authoring files) once a week!

I think what i'm trying to say is that yes, the majority of the data will act more as an archive to be filed away and forgotten, but due to the nature of the business we also need to keep backups of authoring files that are <1 week old.

Sin_Chase, what's a HBA? The agent to run on the desktop?

Ideally, in my mind what we want is a full backup every Friday. I've seen the price of the autoloaders and to be honest I dont think my boss will go for it! That leaves me with a single tape drive and needing to do a full backup once a week, so during the day. Will that kill the network?

Will it be really impossible to set up overnight incremental backups for Mon-Thurs? Bearing in mind the data is all over the place?

If we're talking about a backup plan the interesting thing is that my boss is in Mon+Tues normally, so it'd be ideal to have a full backup for him to take home on Tues evening. Isn't it rather pointless doing a full backup on Friday only to have it sit in the office until Tuesday, where it can go offsite?

Hmn. Lots to think about here!

There's a lot of essays in this thread so i've not read them all through but from the skimming i've done, I can add the following.

A HBA is a Host Bus Adaptor. It's the controller device the tape drive will connect to, which for most LTO scenarios will be an U320 SCSI or SAS host card. You need a good one to get the throughput onto the tape as big as you can so your backups dont run for days. You also need quite fast HDDs for full backups so that data can be read as fast as it can be written to the tape.

Centralising.
You can either re-configure the network to keep all data on one machine or if that's not practical you could use a byte level replication application like rSync to maintain a real time copy of all distributed data in one central place. This in itself if used with snapshots, like volume shadow copy in windows, kind of serves as a backup that's good enough for restoring individual files if someone deletes them or something gets corrupted. rSync is Free or at least cheap depending on what implementation you go for and requires much less work than reconfiguring the way the network is laid out. However it's worth working towards this goal as data duplication will cost in wasted storage space.

Disaster recovery;
This centralised copy and it's snapshots could then be copied to tape and taken off site for safe storage and archiving.
At this point so long as you have a current copy of all the data in one physical place you can take your pick of several ways to get the data from disks to somewhere safe eg, tape, and you can schedule it to run when convenient

Storage needs a holistic approach.

To echo what Skilldilliplop said; Try to centralise all 'served' files and drives into one place. This will make your backup faster, more practical and easier to administer.
I hope that media server is running raid1 at a minimum. Raid5 is preferable for a small file server (Good read speed / drive space utilisation)

Volume Shadow Copy is a nice 'quick' fallback, mainly for people wanting to fish word documents out of accidently deleted folders.
Snapshots are even better.

Once you have centralised your files and put them on a nice RAID, you can start looking at the tape backup.

Tape drives are best hosted on seperate machines to the place the files are being backed, perhaps making one of those desktop machines 'dual purpose' in this regard, with a nice gigabit connection to the server.
Perhaps run Differential backups on Monday / Tuesday, Full Backup on Wednesday, Differentials again on Thursday / Friday. The differentials can be changed to incrementals when size becomes an issue (ie, greater than one tape).

For Backup Exec 12.5 + Windows agent, it will set you back about £800. Keep this in mind when selecting the tape drive and adapter.

Just to add to the above, I have often achieved 1.4 - 1.8TB on an LT04 tape using hw compression and Commvault backup software. This wasn't media files though (Oracle / SQL / Text docs etc)

Well thanks a lot for the help guys. Unfortunately I just spoke to the boss today and we've got our wires crossed. Well, I know what we 'should' be doing/getting but he's not interested. He wanted me to look at getting a single tape drive simply to archive off some of our media in one hit. He's not really interested in setting up any kind of automated backup or anything by the sounds of it. He was looking at single DLT drives for £500, and thinking that will do. It'll be me sitting there for 10hrs to backup half the stuff. Nice.

Do you think that'd count as 'Implementing a tape archiving system and procedure' on my CV? lol.

I was speaking to a reseller on the phone today and he told me DLT was dead. Is this true? Would we be really extremely stupid to get a DLT drive? I can see the tapes are the same price as LTO4s... I'm of the belief that spending more on the drive would be worth it in the long run but I dont think my boss is!

DLT is not dead, it's simply not got the capacity LTO sports.

HP have an 'RDX' solution which is basically Hard-Disks in plastics cases that dock into the drive. I've not played with them yet but for the impressive 60mb/second speeds they have on paper I fear they will not be as resilient as tape. They have a nice pricing point though.

If you are going tape save yourself the headache and go LTO4.

Yes, go LTO. DLT isn't dead but for the amounts of data you could amass in a short time you will want some excess capacity for future proofing.

I would agree, backing up to removable hard disk is very fast but all magnetic disks are very suseptable to G-force and don't age so well due to the moving parts in side. Solid State Disks would be ideal as they don't have moving parts and arn't as suseptable to shocks and magnetic fields. But the Technology is still not there yet in terms of Archive lifetime and GB/£. Infact with SSD it's more £/GB at the moment
Tapes still rule for archives as they're tried and tested and you can buy archive grade ones that are rated for storing data for 15 years +

The problem with LTO's drives are that if you don't future-proof your system, you end up forking out again for bigger backup solutions.
Example: I had a LTO tape drive pre-installed in a new server, 100-200Gb tapes. This catered for our storage back in the day, but as our storage grew so did the need to backup more so our media sets were consisting of 4 tapes and required manual changing of the tapes daily, a real PITA and a very flawed solution.
Then we bought a LTO2 8-tape autoloader, which although uses 200-400Gb tapes, our media sets still consist of 4 tapes. No problem as i put all 4 in the autoloader together and it does a FULL backup overnight. However due to the nature of our files, our compression ratio is approx. 1:1.1-1.3, which aint so good. So we're already nearing the requirement for a 5th tape per media set, which really starts to become a little more unmanageable. So the next step would be to scrap the autoloader and buy LTO4 drive/tapes......where does it end?
External HDD aren't that bad. If they are to be taken off-site daily, surely they'll be put somewhere safe and out of harms way, no?
A 1TB hdd with caddy would cost less than £70. You could buy maybe 7 lots, do a full backup daily for a week and have maybe 2 duplicates of the entire system that don't get overwritten at all.
Just another option really, all backup software will allow for backup to external hdd as much as a tape drive.

The one snag with the externals is the interface. USB and firewire arn't as fast as U320 or SAS Tape drives. So you'd need a SATAII capable e-SATA drive. Most server's don't have e-SATA ports so you'll need a PCI-E SATA hotplug host to cater for it which isn't necessarily practical on a 1u server with only the 1 PCI-E slot. but provided you're not archiving data for long term storage removable HDDs are cheaper in the long run.

The error rate on SATA hard drives (external or otherwise) is at least 100x worse than that of tape drives.

When you're doing a full backup of ~1TB every night, that's quite a few bits of corrupted data.