Deduplication Deployments

I have had the chance to talk to a huge number of customers across the country lately. I love to do this stuff, because: a) I actually really like public speaking, and b) it is a really good chance to level set and get feedback on whether the message is actually speaking to customers' needs, or not.

One of the most startling things that I have found is just how low the percentage of organizations that have installed any sort of data deduplication for their data protection environment. At a guess, based on a polling of maybe 400+ individuals in total, less than 10% of people that raised their hand indicated that they had any sort of data deduplication installed.

A couple of qualifiers here: the mix in the crowd was by no means "EMC Customers" only. Obviously a good number were, but at least half, if not more, were just beginning to have a relationship with EMC. Further, there was also a good mix of both enterprise and SMB customers, and a mix of public and private. In short, I am no statistician, but this seemed like a pretty representative sample to me.

So what did I learn from this? I think we, as an industry, need to continue to be clear and concise on the exact value proposition and message for deduplication. And that is: It saves money. It saves bandwidth (for replication). It is secure and reliable (unlike tape). It reduces data centre footprint, compared to tape. It reduces restore times. It can reduce bandwidth requirements for backup. And it can reduce backup times very significantly. In particular, I think we need to continue to explain that deduplication is less expensive, and therefore has a positive ROI, than traditional backup.

The other obvious lesson is that there are still vast opportunities here. For both vendors and customers. For vendors because as many as 90% of our customers have yet to do anything with respect to deduplication for data protection. For customers, most of you still have a huge opportunity to transform the cost and capability of your data protection infrastructure.

All of which makes for interesting and challenging times for us all.

Comments

[tim]

I have similar numbers, slightly higher, but still far lower than you might expect given the hype. Next year it will be much higher when you ask the same question :)

[W. Curtis Preston]

My numbers for adoption are similar, but they certainly are increasing.

OTOH, I have NOT found dedupe to be cheaper than traditional backup. I think it's BETTER than traditional backup, but I don't see how it can be cheaper. Here are my thoughts and please tell me where I'm in error.

1. The target costs more (a deduped target still costs a little more than a similarly sized tape library filled with tape)
2. The software license costs more. (It costs more to license an intelligent storage device because the ISVs are assigning a higher value to its use.)
3. Dedupe does NOT save money on replication, unless you were already replicating non-deduped backups, which I've never met anyone that does. Dedupe allows you to replace the Iron Mountain truck with replication and makes that replication more affordbable and possible than it would be without replication, but replication is still more expensive than the truck.
4. Dedupe replication requires a system on the other side to receive it, where many people send tapes to Iron Mountain with no unit on the other side. The pay someone like Sungard to have one handy if they ever need it, so they pay for a portion of it. Dedupe and replication requires you to have your own target to replicate to.
5. Many people also want a copy on tape, so we're not even saving money on tape. The whole dedupe system (target in the primary site, license for that, target in the secondary site, replication license, and associated bandwidth) is all incremental cost. How is that cheaper?

I still think it's better, but I never make the argument that it's cheaper.