ATA-Class Storage -- The End of Tape Backups?
DRBC #4: With a solid DRBC plan in place, it might be time to consider replacing your backup plan with more reliable technology that doesn't call for the janitor at a remote site to occasionally change tapes.
- By Bill Heldman
Hey buddy, do you happen to have a floppy? What if someone asked you
such a ridiculous question today? It's absurd, isn't it? You'd probably
reply back: "Where's your flash drive pal? Floppies are so 20th century!"
The same is true of enterprise-class tape backup systems -- only on a
vastly larger scale.In the old days of DRBC we used tape backup devices
and racks upon racks of backup tapes in order to protect our enterprise.
In one (fairly current) job I was in, the operations department still
used five 25-year-old Storage Tek tape backup units. On a monthly basis
they would routinely rotate 5500 tapes. All this for a little over a gigabyte
of mainframe storage! One tape would be used for a little 1MB file, another
for a few hundred kilobytes and so on.
Job security, yes. Practicality, absolutely not.
Think of all of the little one-off 4mm, 8mm, DAT and LDAT tape systems
that shipped with the servers you or someone before you ordered. Administrators
everywhere are still chasing after those tapes! And, people are still
buying robotic tape changing machines to routinely replace the LDAT tapes
in the backup system -- expensive tapes they've procured over the years.
Here's the million dollar DRBC question: Are your tapes routinely going
offsite, using some sort of rotation system? Answer: Doubtful. The sad
truth is that the majority of organizations have elaborate tape back-up
systems but do nothing with the tapes once they've recorded backups onto
them. There is no routine "ship them offsite" mechanism. It
is quite likely that the tapes sit on shelves or in tape containers in
a back room and are not thought of again until the next time they're needed.
Is this good DRBC practice? Of course you know the answer to that. Absolutely
Waiting to Happen
It is not unreasonable to say that many administrators
may have never even run a restoration test on their
tape backups to see if they can actually restore a file
This happened to me one time. I was using Arcserve
over Windows NT (yes, I'm dating myself). I saw in the
tape backup administration software that I had successfully
backed up files -- at least I thought I had. But I never
practiced actually restoring the files to make sure
that I could do it -- or at the least, to work out the
details of exactly how to restore the files, should
I need to. Instead, I just naively assumed that since
I'd validated the files were out there on tape, they
were ripe for the taking.
Not so! On the fateful day when I actually needed to
perform a SQL Server database restoration, guess what?
Even though the tape back-up software said the files
were available, there was no way I could persuade the
restoration to actually do its thing. The result? Many
hours of lost time recreating the database. Fortunately,
there wasn't much data in it yet. What if it had been
loaded with data? I would've probably been let go.
"But", you say, "I can't give up the incredibly expensive
investment I have in my tape back-up units. And what about the back-up
software I use? That software cost us an arm and a leg! You've gone too
far. You're talking a major capital investment in software and hardware,
not to mention an Operations and Maintenance investment in professional
services to get it installed and operational."
True. You do have a capital investment in your tape back-up gear. But,
realistically speaking, how long has it been in service? If the answer
is two years or more, odds are you are close to the end of the depreciation
cycle for that equipment and it's time to venture into greener pastures.
You should begin to think of purchasing ATA-class disk array equipment
to replace your backup units. We'll talk more about these devices later
The software you'll likely be able to keep, as it will probably be able
to converse with today's ATA disk array equipment. And if not, there is
software available that will make it easy for you to transport data from
your servers to the disk array without using tape backup software. But
for the time being keep "I'll probably be able to leverage my old
tape back-up software across the new ATA-class disk arrays" in the
back of your mind for later.
The bottom line is that the new breed of ATA-class storage arrays are
relatively inexpensive, they have awesome companion software available
for them, they are manufactured by world-class manufacturers, and they
provide you with a safe way to produce backups and restorations much more
quickly and reliably than before. Plus, they juxtapose nicely with DRBC
You want ATA-class back-up operations because the speed with which you
can do your backups is no longer limited by the groggy speed of the tapes.
Now your primary bottleneck will be your network bandwidth, and we all
know that today's networks are usually quite robust.
Used to be that performing a backup was an overnight thing. Now you can
plan your backups to fit your daytime schedule. Restores were a problem
because they required navigating the software until you found the volume
you wanted, then actually finding the tape or tapes that you needed in
order to retrieve the data. This slow, painful operation is now quick
and much easier than before.
Let's talk about some of the manufacturers -- folks you may have
already dealt with. You may even possess some of their equipment. Please
note that I've just listed three of my favorite players. There are many
others to choose from.
Network Appliance -- This
company has an amazing lineup of SAN and NAS gear for almost any undertaking.
This company's products work well, and the company stands behind them.
The NearStore Virtual Tape Library (VTL) is a disk-to-disk SATA-based
backup appliance that appears as a tape library to your backup software.
Available in capacities from 4.5 Terabytes (TB) to 168TB, there is plenty
of room for you to back up old data and get it off of your servers. The
VTL is visible by today's enterprise-class backup products such as Legato,
Symantec, HP, et al.
If you've been hankering to get rid of your clunky old tape backup software
program, you can combine the VTL with NetApp's SnapVault software. This
combination immediately pops you into a modern disk-based backup paradigm
that allows you to reach out and quickly back up anything on the net.
NetApp also has other enterprise-class SAN and NAS arrays that can be
utilized in conjunction with the NearStore. Architecturally, you can probably
envision SAN arrays sitting at your organization's main points of operation,
fanning out offsite to one or more NearStores that pull in data from the
enterprise and store it out of harm's way, all over a fast wire. No more
hauling tapes around (which few of us did anyway), or worrying about Father,
Son, Grandfather or Towers of Hanoi backup strategies.
LeftHand Networks --
LeftHand is in the business of iSCSI SANs. Because they're iSCSI, you
don't have to mess around with complicated Fibre Channel connections,
and the equipment is thus less expensive to operate. Talk to the big players
like EMC and you might hear stories about iSCSI not being standardized
across vendors -- scary interoperability stories. That's because the big
players have a huge investment in FC arrays. But the truth is, iSCSI works
as a standard and LeftHand has been shipping gear for a long time now.
LeftHand has some software called "SAN iQ/Snap". The key word
here is "snap", meaning, more or less, a snapshot of the data
that you want to back up. The idea is straightforward: You designate the
files that you want to snap and allow iQ/Snap to do its work. Note that
these files are on the SAN, not necessarily on a local server. Once the
snap has been taken, it's an easy thing to move it to another device out
of harm's way. Note that iQ/Snap will work with some implementations of
Windows-based backup software.
Snaps typically require a dedicated piece of space on the SAN in which
to hold the snap. Not so in LeftHand's case, which can place the snapshot
LeftHand is a Microsoft Certified Partner, and this shows up in their
products. Note that alongside iQ/Snap, LeftHand has an implementation
of Microsoft's Volume Shadow Copy Service (VSS), allowing your server
VSS implementations to initiate and manage snaps from the LeftHand SAN,
ostensibly for movement elsewhere. Additionally, LeftHand has leveraged
Microsoft's Multipath I/O (MPIO) framework to optimize the ability of
a Windows Server 2003 server to work with SAN data. LeftHand's SAN iQ/Remote
Copy software allows you to pick up data from remote locations, copying
it to the SAN for safekeeping. This would be tremendously useful in those
locations where you had no administrator and wanted to be sure you were
adequately and accurately backing up the site.
It seems to me that the LeftHand implementations are proprietary, single-stop-shopping
environments. This might mean extra effort in getting a LeftHand device
to "play nicely in the sandbox" with another device on the network.
Be that as it may, the upsides are the cost-effectiveness of the software
and hardware -- all wrapped up in an iSCSI framework.
in a Day's Work
In another not so long ago job, we had a little over
50 geographically disparate sites -- single offices,
really -- with just one or two people in them. We had
locations in Chicago, Atlanta, Houston and so forth.
Each site had a server, with a little tape drive in
it, and no dedicated administrator. Instead, we had
the sites broken out into four or five regions, each
of which had assigned to it what I'd basically call
a PC Technician on steroids. This person would routinely
fly around to all of his sites, solving user problems,
helping prepare reports, and, most importantly, managing
those all-important backups.
You can imagine how easily the backups did not happen
once he left the office and things settled down
for a week or two. This PC Technician was expecting
the office manager, or the administrative assistant
(in one case, the janitor) to rotate the nightly tapes.
You can guess how effective the process was.
We accessed the servers remotely with Microsoft Systems
Management Server Remote Tools (this was a year or two
before RDP was available). One time I accessed a site
and was startled to see the banner of a CD-based game
on the screen! Evidently the office manager decided
he wanted to install this game to see how it worked
and left the CD in the caddy at the installation banner.
Fortunately, he or she did not get the opportunity to
go through with the installation, but it was freaky
to think that someone hired to sell product was able
to nonchalantly walk into a server closet and insert
Now imagine that you have a snap copy software product
that can reach out and touch these volumes, quickly
grabbing a snapshot of each for storage on the disk
array. How much easier would it be to restore the server
if that manager had, for instance, gone through with
the game installation and caused a BSOD on the machine?
EMC -- One of my favorite companies
is EMC. I take a ribbing from a lot of people; I suppose because EMC is
like Microsoft, one of those companies everyone loves to hate. But my
experience has been that all EMC people I've been in contact with -- salespersons,
technical reps, installers and service personnel -- have been top-grade.
I like the products, and I like the way I'm treated when I'm an EMC customer.
EMC's ATA-class box is called the Clariion. The flagship backup and restoration
appliance is currently the Clariion disk library series. Using a device
in this series allows you guaranteed interoperability with other elements
of your network that need backed up, and provides you with a path from
Small to Medium Business (SMB), all the way up to large-scale enterprises.
EMC's software for backups is called EMC Networker. There is a Networker
module for most enterprise ERP, database and email implementations.
EMC has an Insignia line of software and hardware designed for SMBs.
So How Do I Start?
Clearly, these are just three examples in a universe of vendors
who are in the backup-to-disk market. The technology is tried and proven.
It works, and it gives time back to administrators -- most of whom typically
spend their days as though their hair is on fire. It is time to consider,
architect, plan and implement a backup-to-disk scenario for your DRBC
So how do you start? Well, I think the best place is at the top, with
the CIO. You need to understand four basic things in order to prepare
a document that's going to get you what you want:
Outline the need with statistics -- Most administrators try to
sell their managers on products by simply insisting that this is what
they absolutely must have to continue operation. Folks, let me be frank.
Your CIO didn't just fall off of a load of pumpkins. She knows that up
till now, you've gotten along just fine without all the latest and greatest
software and hardware. And, she has a variety of hands out, all begging
What you need to do is document the time and money that is wasted using
current methodologies. You can show how things have gotten busier over
time -- where once the situation might've been OK, now it is a nightmare.
You must illustrate to your CIO how the situation has gotten out of hand
so that she clearly understands the problem you face.
Understand and talk about capitalization -- Once you've prepared
a sound problem statement -- one that is going to get noticed and get
some airplay -- the next thing to do is to make sure that you understand
how big equipment is capitalized. You must understand financing, depreciation,
the projected length of operation of the new equipment, the project plan
needed to implement the system, and any additional costs that are needed
to install it.
You want to be able to illustrate to your CIO why it's OK to bag those
old tape back-up devices, and how you propose financing the new backup-to-disk
arrays and software. You want to tell her how long you anticipate keeping
the devices, because this denotes the depreciation value of the equipment.
Also, it would be wise to point out anticipated Mean Time Between Failures
(MTBF), as this illustrates any expected downtime you might encounter.
By simply talking in capitalization terms, your CIO grasps that you basically
understand what she is up against when it comes to budget planning and
Understand and talk about O&M -- Big-league gear like this
doesn't come as a standalone box. There are ongoing maintenance costs
you have to pay for things like 24x7x365 monitoring, as well as software
upgrades. You should be able to talk about the operations costs -- those
people in your shop that will actively monitor and maintain the equipment,
so that she understands what today's gear looks like as contrasted with
what you're proposing. It could be that full-time employees (FTE) that
are currently working backup operations could be redeployed elsewhere,
thanks to the deployment of the new solution.
By putting things into a dollars and cents mentality, it will help you
rationalize the expenditure you're proposing. She will realize that you
are trying to think like she thinks, and not just demanding that you must
have the latest and greatest thing.
Risk assessment -- You also want to provide a full risk assessment.
You start by outlining the risks that the organization faces today, given
the current situation. Then you move into how the new implementation will
reduce those risks. You should also talk about the risks you expect to
encounter with the new proposed solution. Don't be a "pie-in-the-sky-guy".
There will be risks that arise with the new system. You should spend some
time thinking about what they are and documenting a mitigation strategy.
Note how this new solution fits into your over all DRBC architecture
-- By this time, you've doubtless been dialoging with your CIO regarding
DRBC. Your segue moment comes when you can show how the current tape backup
paradigm really isn't DRBC -- if anything it's a loose fit disguised as
DRBC. And you can point out how the new system will map very nicely into
your DRBC architecture because one or more of the disk arrays will live
elsewhere, away from headquarters.
When you're done preparing your presentation, schedule an appointment,
meet with your CIO and other dignitaries and make your case. You start
with why the current situation is no longer palpable, given the current
infrastructure. Next you go into the research that you did with regard
to the latest solutions -- including a discussion of acceptable industry
standards. Finally, you introduce your proposal, including your risk assessment
and end with the proposed cost of the new solution, along with financing
plans, and DRBC architecture fit.
Then you ask for the dough.
my second article in this series, I talked about
Risk Management, and said that it consists of: Analysis,
Assessment and Mitigation. What I neglected to mention
is that one form of Mitigation is something we call
What this means is that you recognize a risk and decide
you won't fool with it. You'll outsource the problem
to someone else.
Example: You determine that you need a DRBC
offsite location. After assessing build-out costs and
all of the other headaches, you determine that your
friendly local offsite vendor could just as easily handle
the load (for a fee, of course), offloading you from
the risk of maintaining a DRBC offsite location.
Risk Transfer is a very viable choice to make when
going through a risk assessment methodology. Don't rule
it out at any stage, any risk.
The point to this is that if you have not already replaced your current
tape backup system, why not? There is literally no need for such gear
in the enterprise any longer. Disks are fast and reliable, tapes are not.
You can take snapshots all day long, and when you are finally ready to
get rid of very old data, you can simply burn it to DVD for safe, long-term