Backup Isn't Dead, It's Just Evolving

The blogosphere can be a crowded, noisy place -- especially when it comes to storage-related topics.  Sometimes I think people say outrageous things intentionally. 

I suppose if your goal is to draw attention to yourself, the tactic might make sense.

But if your goal is to help others, saying outrageous things just adds to the noise and confusion -- especially around a relatively important topics such as backup -- and really doesn't help anyone.

One of the best examples of this phenomenon in storage land is the recently popular "backup Is dead" meme.

On the face of it, it's simply another attention-getting claim.  But upon deeper consideration, there's a bit of merit to the thought.

Certainly, backup -- as we've historically known it -- is undergoing substantial change, but that doesn't mean it's "dead". 

Caterpillars turn into butterflies -- but they don't die in the process.  

What Do You Mean By "Backup"?

I've always thought of "backup" as the act of creating recoverable data sets in the event that bad things happen.

"Bad things" include hardware and software failure, logical corruption of information -- all the way up to the proverbial "smoking crater" scenario.  You want to create a protected view of how information looked at a given particular time -- and put your copies as far away as possible as you can justify.

Experts in the discipline will tell you backup and related disciplines can be an incredibly sophisticated topic: discovering information sources, classifying them by importance, understanding the relationships between applications, scheduling the backup process, ensuring it worked, practicing recoveries, doing audits, migrating to new technologies, etc.

There's a lot to it.

And, given that (a) information is growing at about 60% annually, (b) the importance of information is generally increasing, and (c) the consequences of losing data are increasing -- well, it's just hard for me to avoid a visceral negative reaction to the statement "backup is dead".

If anything, the need to intelligently protect information against loss or corruption is growing exponentially.

The Underlying Technology Is Changing, Though ...

Sure, tape has always been the historical medium for backups.  And I could make a strong case that using tape for backup -- while not technically "dead" -- is certainly on extended life support.

The preferred medium for stateful copies is now disk.

Between dropping media costs, exploding data volumes, enabling technology such as data deduplication, and the need to recover faster, it's the rare IT shop that isn't moving away from tape and towards disk for their future needs.

So, I could live with the statement "Traditional Backup Media Is Dead" ... but that's not the claim.  The claim is that "replication replaces backup". 

And I've got an issue with that.

It's All About Copying Data

I could argue that the act of performing a backup -- to any media -- is another form of replication.  You're essentially making a copy of data that was living in one place, and putting it in another.

Historically, the fact that tape had different properties than disk necessitated a change in format -- from "disk volumes" to "tape volumes" in essence.  And one could argue persuasively that -- in a world where we're backing up to disk -- this historical format change isn't always needed.

And we've all known that time-to-recover is far faster if we happen to have a friendly copy of the needed data sitting on disk in native format.  The copy can potentially be used for other purposes as well.  Indeed, making what appears to be native copies can be potentially better than going through a no-longer-needed format change.

But there's far more to backup than simply copying data from one place to another.

You still have to discover and classify different information sources.  You still need to understand the relationships between applications.  You still need to ensure that the copies are being made on schedule, and that they're recoverable.  You still need to coordinate with applications to get a consistent view of the information you're saving.  And you still need to create a catalog of saved images, and manage them effectively. 

Not to mention being able to monitor the whole shebang effectively.

Whether the data is moving from disk to disk or disk to tape doesn't make any of this other "backup" work go away.

Thinking Of Information Protection As A Service

If you agree with the thought that IT is evolving to an "IT as a service" model, perhaps the most useful construct is "information protection as a service". 

On one side the equation, you've got your service catalog -- what you deliver to the business.  As you construct your service catalog, you're going to want to deliver a variety of information protection service levels based on what's needed, and how much budget you've got.

On the other side of the equation, a variety of protection mechanisms: backup, archiving, replication, etc. -- usually from a variety of vendors, EMC included.  There's newer technology rolling in, and older technology rolling out.

In between these two, you need the ability to deliver information protection as a service, much as an external service provider would do.

To do this, you've got to be able to do a lot: discover information sources, classify them, understand their relationships, invoke different protection mechanisms, report back on their success, monitor resource usage and elapsed time, and -- basically -- be able to say "yep, I can recover this data within the agreed time".

That's why I think that service management products -- like EMC's Data Protection Advisor -- more accurately represents the new wave in information protection.  Technologies like these evolve the discussion from "backup" to "information protection as a service".

And whether you're doing traditional tape-oriented backup, continuous data protection to multiple sites, or anything in between, you're in control.  The best part?  You can easily create portals that let users inspect and verify that their information is protected and recoverable -- just like a service provider would do it.

Separating The "What" From The "How"

So often in this industry, we confusingly tangle "what needs to be done" with "how it's done", and the current discussion is no exception.

The need to protect and recover ever-growing amounts of information (e.g. backup) isn't going away any time in the foreseeable future.  To claim otherwise is simply ridiculous, IMHO.

But the "how" of all of this -- using disk vs. tape, making copies in more accessible and usable formats, moving to a service delivery model vs. point solutions, increased use of external services, etc. -- well, that's certainly changing -- and fast.

We can only hope our homely caterpillar can morph into a beautiful butterfly.

Comments

[Mike_fishman]

Chuck: This is a great piece. I recently presented a related talk at SNW 2010 entitled "Introduction to Data Protection: Backup to Tape, Disk and beyond". Your points about backup media are spot-on. Tape is not quite dead yet.

While this backup blog post was interesting, I think you'll grab more interest if you talk about the challenges and implications of recovery.

[Rob]

I don't think this message can be posted often enough. The poster child for what you talk about is JournalSpace.com. They relied on RAID and mirroring as the soul means of protecting the data but when the data was wiped out on the source, fast replication promptly wiped it out on the target. They didn't go out of business but they don't have the same owners today. My guess would be that the original owners got a soaker and the new owners have an uphill battle of re-establishing JournalSpace's credibility. Of course all the other IT shops in the world learned their lessons and invested in a proper backup strategy......(?)

[Christoph]

Let me start my comment with my favourite outrageous(?) statement: Nobody wants Backups!!
Backups are a typically error prone process, interfere with production, keep IT personell busy and therefore are associated with "trouble".
What people really want, is the ability to restore/recover lost data and lost application services. If you start from this perspective (i.e. with looking at the requirements first) you'll quickly see weaknesses of many backup implementations.
You mentioned some of the typical weaknesses and also some means to deal with them. I totally agree that tools like Data Protection Advisor (DPA) are great to measure and monitor Backup (or Replication ;-)) and Restore processes.
Nevertheless: tools like DPA tell you things about the "known world" only, i.e. about environments you run backups for. But how do you ensure that you don't forget to include a server or database in the backup at all?
We ran a relatively simple test at a customer site with a combination of SMARTS Application Discovery Manager (ADM) and DPA. ADM passively discovers all servers and services (and even their dependencies) from listening to the network traffic. We fed its output into DPA and reported servers/services in the network against those found in the backups.
We found quite a number of servers and even databases that were never backed up.
Management Tools are essential in a quickly changing and more and more complex environment. But make sure to check for "uncharted territory". This is essential, as well, to stay in control. Even more during rapid changes.

[Dominic Cody]

Backup always seems to be a hot topic for a while then vanish then come back to bite people. Over the past 4 years I would say the industry take on backups has really driven forward.
Of course companies and even the techy people themselves get caught up on the idea of backups and archiving and can't actually seem to distinguish between the 2. They are completely different things and people need to start to understand this.
The one thing that is true beyond any doubt is the fact that backup is changing in the first instance moving from tape to disk/VTL with De-Dupe but then of course tape always needs to be involved. And then Archive does need to be linked hand in hand as this could be a better course of action for some data rather than backup. It all boils down to a user case by case basis.

Does that make any sense (not sure myself now)

Dominic

[LTO 4 Tape]

that's a good definition... thanks Chuck!

[Brian Kevin Johnston]

Thanks.. Great article... "Thinking Of Information Protection As A Service" I could not agree more.... best, Brian-