De-Dupe Vendors Come Clean on Marketing Claims

While storage vendors in general are notoriously glib, the area of deduplication offers so much grist for the marketing mill that getting the straight scoop on products can be a challenge.

Two popular de-duplication suppliers -- Data Domain and Quantum -- today clarified key points over which hype has led to confusion.

In a conference call this afternoon, Byte and Switch questioned Data Domain spokespeople about how they can call their DDX series an "array" when the DDX does not yet support deduplication across multiple storage nodes.

Indeed, Data Domain CEO Frank Slootman told us in January that this kind of clustering was not part of the company's products yet, but is on the roadmap for 2008.

Data Domain VP of product management Brian Biles says the company means "array" in the sense of being able to manage up to 16 controllers in one box, not in the sense of de-duplicating across multiple attached backup devices. "We are very clear in our materials. We're not trying to fool anybody," he says.On to Quantum, where we posed questions about the capabilities of that vendor's deduplication appliances, specifically the DXi7500, which was announced

in June 2007 but has yet to be generally available.

The vendor addressed an observation on this publication's message board that the new box only performs inline deduplication at low ingestion rates, implying that the unit is mainly a post-processing device. Quantum has always billed its DXi series as capable of deduplicating via both inline processing (as data is sent to backup) and post-processing (after data is backed up to disk).

According to Quantum CTO Jeffrey Tofano, the DXi7500, which has been sold to select customers and is set for general availability later this quarter, indeed supports both inline and post processing. Customers can make the changeover dependent on specific criteria.

Some background is in order here: Quantum supports an "adaptive engine" across all its DXi deduplication appliances. The units perform inline deduplication until the speed of incoming data would either force the machine to throttle back or start writing to disk. At that point, the appliances, which use their own internal monitoring to discern data rates, start writing to disk, using the post-processing approach.

The point at which post-processing starts is a function not just of data ingestion but of the speed at which the unit is reclaiming space and performing data replication for backup on attached gear, Tofano says. The inflection point varies with application.For example, links between a DXi and an NAS are typically slower than links between a DXi and a VTL. In general, Tofano says, NAS rates cap out at about 200 Mbytes/second, whereas VTL rates can soar to 1,200 Mbytes/second.

Quantum has designed each DXi with a specific application in mind. The DXi3500, for instance, is commonly used only for NAS connections, and Quantum guarantees a steady state throughput of 125 Mbytes/second. There is typically no situation in which the box would have to click over to post processing, since that shift doesn't usually occur until rates are in the 200 Mbytes/second range.

The DXi5500, in contrast, will support rates of up to 250 Mbytes/second for combined data ingestion, replication, and internal space reclamation. This means it can be used for NAS connections at up to 200 Mbytes/second, and then it will switch automatically to post processing for VTLs.

On the upcoming DXi7500, users are given a choice to set the shift from inline to happen according to policy. They can order the machine to shift to post processing for NAS, for example, instead of automatically using inline processing.

The analyst who made the message-board inquiry about the DXi7500 thinks the explanations are fine, but he urges prospective buyers to pay attention to the fine print and to ask questions when something isn't clear. "It's important that everyone knows what a given product really does," says W. Curtis Preston, VP of data protection at GlassHouse Technologies. "Unfortunately, that doesn't always match what a salesperson says it does."Have a comment on this story? Please click "Discuss" below. If you'd like to contact Byte and Switch's editors directly, send us a message.