By "information protection", I mean things like backup, recovery, replication, etc.
For the last few years, customers have begun to get comfortable with the idea of "tiered storage" -- put the right information on the right service level at the right cost at the right time.
Simple in statement, but potentially hard without a bit of forethought.
I'd offer that -- during 2008 -- we'll see a strong interest in "tiered protection" -- protect the right information with the right service level at the right cost at the right time.
And, despite all the new technologies flooding into this space, it might end up being easier than anyone might have thought.
Basic Tradeoffs
Anyone who's got more than a passing familiarity with storage knows there's a wide range of service levels and costs.
At the low end, we've got things like 1TB drives that might be used unprotected. And at the high end, we've got very fast/expensive disk drives that can be short-stroked, as well as double (or even triple) mirrored.
Lots of ground between the endpoints, and it's getting ever-broader. Match the right information to the right storage, and you'll save a bunch of money, especially at scale.
The same sort of tradeoffs exist in classic information protection.
Using disk as a target instead of tape. Local and remote replication. Combinations of approaches.
Big ranges in costs and service levels -- everything from "we're not backing it up" to "we have multiple, synchronized global copies".
Just as there's no "perfect" approach to storing information, there's no "perfect" approach to making safe copies, either.
Lots of ground between the endpoints, and it's getting ever-broader. Match the right protection to the right information, and you'll save a bunch of money, especially at scale.
Just like it did with storage, this leads you to the notion of a service level catalog: here are the services you can provide for information protection, here are the tradeoffs, and here's what they cost.
And, for the last few years, we've been working with many customers to define just that sort of service level catalog approach for backup and replication.
If our experience with tiered storage is any predictor, you're probably looking at somewhere between 20% and 35% annual cost savings, just by creating tiers and exposing true costs to the business.
I'd expect greater savings here, because the ineffeciences are much greater.
And There Are Some New Choices
We've always had tape in different flavors.
We now have virtual tape libraries and disk libraries in different flavors.
There are several distinct flavors of local replication (snaps, clones), each with distinct tradeoffs.
Sure, there's been remote replication for a while, with different modes (sync, async, point-in-time) but CDP (continuous data protection) is getting increasingly popular.
And, just to top it all off, there's now at least a half-dozen flavors of data reduction technology in the market (sometimes called data dedupe): client-side, target-side, single-instancing, etc. etc. that present even more choices.
Three Safe Predictions
Prediction #1: You are most likely to be interested in multiple protection approaches over time
Even the die-hard tape users are starting to use other forms of information protection. Ditto for the die-hard array-based replication crowd.
I can't believe that any shop of reasonable size wouldn't have a need for two or more of these approaches in their environment.
Prediction #2: There will be more choices to make copies of information in the future, and not less.
The pace of innovation in this space has been breathtaking over the last few years, and I don't see any sign of it slowing down. We'll see dedupe everywhere in the stack, new forms of replication, journalling, etc.
Prediction #3: The more different stuff you put in your environment, the harder it may be to manage and control it all.
You've probably seen this movie before. Buy a bunch of stuff that seems "optimized" for your environment, and then you spend too much time trying to make it all work.
Towards A Universal Client -- And Universal Management?
One way of taming the complexity beast is to think in terms of a universal backup client -- a single agent that can target just about anything you'd like ...
Tape. Disk. Disk pretending to be tape. Snaps. Clones. Remote sync/async/point-in-time/CDP. Dedupe client. Dedupe target. Maybe even a laser printer (just joking!)
You name it, you can point to it.
No one wants to rip-and-replace what they've got. But a universal client goes a long way to minimizing that pain. As different candidate technologies become attractive, or as needs change, it's a simple matter to simply "point" the universal client at a new backup target.
Disk instead of tape. Dedupe instead of incremental. Clones instead of snaps. And so on.
Doesn't eliminate complexity, but can potentially take a big bite out of it.
A while back, I wrote about the new Networker functionality, and the one thing that jumped out at me was the "universality" of its backup client.
Not only could a single client target everything in the EMC portfolio (insert ridiculously long list here), but it does a good job of targeting things in other vendor's portfolio (insert second ridiculously long list here).
Not perfect, but pretty good, I'd offer.
And, of course, Networker comes with a backup management system that knows what's been done, where the copies are, and (best of all) how to bring them back if you're in a hurry.
Regardless of target.
I would offer that's a useful abstraction, especially in a tiered protection environment -- one where you're trying to match information protection requirements with technologies, and the trend is towards more choices, and not fewer.
So, When Is The Industry Discussion Going To Change?
In some ways, we're worse than gossiping schoolgirls when it comes to new technologies. We talk in terms of the "best" technology from one vendor or another.
As if the new stuff will completely replace the old stuff.
What really happens is that another option joins the party. Customers have to figure out how they use the new stuff alongside the old stuff, and not go crazy in the process.
I'm hoping that -- sooner or later -- we end up with an industry taxonomy that helps customers figure out where one protection technology or another fits in terms of service levels.
I think it's inevitable that most shops will end up either formally or informally end up with a tiered information protection environment. The stakes are too high to be able to say "we'll only do it this one way".
So, will we be talking about "tiered protection environments" and "universal backup clients" in 2008?
I hope so -- it's about time.
Comments