There’s a tendency among marketing types to squeeze a new product into a single big feature (headline marketing). I understand, I’m a marketing guy myself. But there are many times when the value of a product really comes from more than any one thing—and it usually means that the product is much more interesting than can be described in a headline. The new DXi6700 solution that Quantum has just introduced is a great example.
Development of the DXi6700 was driven by discussions with IT departments that either haven’t used deduplication at all, or that have tried it but found it too limiting. What were the issues they raised? For one, people said that dedupe vendors set up too many limitations when during the initial decision—these businesses just didn’t know what they were going to need in the future when their backup needs changed. They weren’t sure what capacity they would need, or how fast the systems would really go. They weren’t sure what kind of interconnect they would need, or if they wanted to stay with a target appliance or do dedupe in software. They were also unhappy about prices and about “surprise” costs—usually extra software licenses—that showed up later.
The new DXi6700 models—the DXi6701 and DXi6702—were built as a way of keeping people from having to limit themselves by making too many of those compromises at the initial point of decision, and as way of reducing overall costs. To do that, we designed the basic features to enable easy changes over time. The backup performance boasts up to a sizzling 5.8 TB/hour (I still think in terms of MB/sec—it’s just over 1610 of those), so it’s fast enough to take care of even very demanding backup needs (it’s far and away the fastest in its class).
In addition, it’s not limited to specific backup software or a single interface. For size, the DXi6700 models scale all the way from 8 to 80 TB—a huge range for a single appliance—and capacity growth is simple. For interface and presentations, all the options are standard: CIFS, NFS, VTL, OST, and both 1GbE and 10GbE interfaces are supported, along with FC. So, whatever the business ends up needing is there.
To help address the objection over costs, the MSRP of the units has been reduced to less than earlier generation systems. And all the software licenses are included in the base price—that includes all interface options, replication, even Quantum’s direct path-to-tape feature. So the value is extremely high, and the return on investment window is reduced to less than a year.
The last big change evident in the DXi6700 is Quantum’s introduction of new software—DXi Accent. This software extends deduplication out into the backup infrastructure, which makes backup faster over low bandwidth networks and gives users what may be the first practical system for a software-only alternative to target backup devices in remote offices. The DXi Accent story is interesting enough that I plan to devote a future blog entry to it, so watch this space.
For more on the new DXi6700, visit: http://qntm.co/DXi6701DXi6702

will it support arbitrary 3rd party storage as target? The $/TB for these appliances is far beyond obscene. I don’t mind buying the head unit and the software and intelligence that go with. But I want to hook up arbitrary SAS/FC/iSCSI-presented LUNs to the unit. Think budget storage enclosures the like of BackBlaze or HP MSA600 or DDN/NexSan bulk expansion chassis.
We also realize that cost is important. You might be surprise at the pricing of the new DXi6700 units–they are substantially lower than earlier generations, and about half the cost of competitors. Users also tell us that they save enough money by putting dedupe appliance that the systems pay for themselves in less than a year–pay attention to operating expense as well as acquisition.
Technically, putting a dedupe head on generic storage can work, but it usaully means substantial compromises in performance and utility–more than we think makes sense for most customers. It also requires significantly increased qualification testing which raises costs. That’s becuase to work well, the disk needs specific RAID configuration, the data integrity features of the disk subsystems need to integrate with the appliance itegrity protocols, reporting features need to support detailed, disk-level details, and full regression testing has to be maintained across multiple applications and oerating scenarios. Those issues don’t mean that this product approach won’t ever make sense, but it helps to explain why several vendors who started down that path ended up focusing instead on providing more traditional appliances.