Deduplication Guarantees do not Guarantee Money in the Bank

Are deduplication guarantees really something you can take to the bank? As more companies look towards using disk in general as a backup target and deduplicating systems specifically, deduplication guarantees are emerging as a way to influence users' decision to deploy deduplicating systems. But in these tightening economic times, deduplication guarantees do not necessarily guarantee money in the bank and may shift your attention away from more critical evaluation criteria such as system reliability, scalability, and performance.

On the surface, deduplication guarantees of 20:1 or higher may sound like a good idea. Assuming recent surveys are accurate, most users have not yet implemented deduplication in production but plan to do so over the next couple of years. The lure of a guarantee is that it can take some of the uncertainty out of making the decision to deduplicate backup data by providing some re-assurance to the user that if the system does not deduplicate as promised, the vendor compensates them for the shortfall.

But when one takes a deeper look at the deduplication guarantees currently available, holes in the terms of these guarantees become more evident. For instance:

• The guarantees are only applicable to specific applications that satisfy certain conditions. For instance, unless companies plan to backup and deduplicate data from Exchange or Oracle a specified number of times per week, the deduplication guarantee may not apply.
• The vendor is the final arbitrator as to if the conditions of the guarantee are met. You need to use the vendor's software to measure the results and then the vendor interprets the results. This is not to imply the vendor will lie but it does raise some questions as to the validity of the results.
• Remuneration is vague. If the vendor fails to meet the terms of the guarantee, the compensation may consist of additional hardware/software or services from the vendor. Yet in reading through the terms of the various guarantees on the market, the associated corrective actions are vague. For instance, if you need more storage capacity because the guarantee was 20:1 and you only achieved 15:1, how does that translate into software, hardware, or services that the vendor will provide?

So if deduplication guarantees are not the answer, what is? A better approach is to deal with vendors that are willing to bring their solutions into your environments and then let you test them to see how they actually perform under real-world conditions. NEC offers companies that option with its HYDRAstor. NEC will provide qualified companies the opportunity to run tests with an appropriately sized, production ready HYDRAstor. The HYDRAstor is sized to match the needs of that company's environment that the company can deploy, run for the needed period of time, get actual deduplication results and then make an informed decision about how to best proceed with the NEC HYDRAstor since they can see the real cost savings that deduplication provides. This approach will also allow you to evaluate more critical aspects of the system such as reliability, scalability, and performance, to ensure the system can address both current and future requirements.

Deduplication guarantees may provide some short term assurances but are not money in the bank as they are for now still too narrow in terms of the benefits they offer. Even if they offer a full return policy, very few organizations can afford the disruption and churn of replacing a production system due to failure to deliver on a deduplication guarantee. While there may eventually be a time and a place for deduplication guarantees, the conditions that current deduplication guarantees attempt to satisfy are not reflective of the reality of most customer environments. Instead, companies are better served by choosing solutions such as the NEC HYDRAstor that give companies the opportunity to prove the ROI while avoiding the need for them to make a leap of faith that a deduplication guarantee is intended to provide.