It is hard to believe that it is October 2009 already and that we are over a year removed from last year’s economic meltdown even though everyone keeps talking about it like it just happened yesterday. However, the upside of the situation that we are in right now is that it is prompting more organizations to think outside of the box and look for better ways to do things.This is resulting in companies choosing some technologies that they may have overlooked or not considered in the past.
A good example of this is an IT manager I spoke to a little over week ago. He works for a rapidly growing radiology practice in the Northwestern United States and needed to upgrade his storage infrastructure to support all of the new physicians and hospitals that it was supporting. He recognized that a networked storage solution was the right answer to his set of storage requirements but his budget did not allow him to purchase a high-end million dollar storage solution so he sought out a midrange storage solution that met his specific needs.
Initially he was attracted to midrange solutions from EMC and Dell as they appeared to provide the features that fell within the scope of his budget but when he examined the underlying architecture of these storage solutions he uncovered some issues. Specifically, both of these solutions included many redundant hardware components to account for hardware failures but he still found instances where individual points of failure existed within these systems.
His other concern had to do with maintaining performance and availability while performing routine maintenance on these systems. Here again, he saw problems with these solutions. Not only did these midrange storage systems experience degraded performance of 50% or more during firmware upgrades but if something unexpected happened, it could result in a complete system outage which was unacceptable in his environment.
This led to took a closer look at a lesser known storage system: the Celeros EzSANFiler XD storage system. While it had an attractive price point (he could purchase four of the Celeros systems for the same price as one Dell or EMC system and get up to four times as much storage capacity), it did not have the availability or hardware redundancy that met his business requirements.
However what changed his mind was Celeros’ synchronous replication feature that mirrors data between two Celeros storage systems. Deploying the Celeros systems as pairs in a mirrored configuration, he could create a solution that delivered the availability that his business needed at a price point it could afford.
Another example of an issue that is forcing organizations to look outside the box for new solutions is the need to replicate data backed up to local storage devices to a central repository. As organizations more aggressively deploy disk-based deduplicating backup appliances in more remote and branch offices, it creates new needs to replicate the data from these remote sites to a central production site and then from the central production site to an offsite DR site. Sounds logical but try to find a solution that actually does that for these size environments, especially when one begins to look at large fan-in ratios and cascaded replication.
These new data protection scenarios are part of the impetus behind last week’s DD OS 4.7 announcement from EMC for its Data Domain inline deduplication storage. DD OS 4.7 now supports fan-in replication ratios as high as 180:1 (i.e. – 180 branch offices with a Data Domain appliance can replicate data to a single central Data Domain DD880 solution) and can do cascaded replication. Once data is replicated to that central DD880, it can then in turn replicate that data offsite to another solution and has been tested to as many as 4 hops.
To say that Data Domain’s customers are not interested in replicating data offsite would be understatement. Data Domain has been monitoring replication attach rates on its units for a couple of years now and finds that 60 – 70% of its units each quarter ship with replication with attach rates as high as 95% on some units that support specific protocols.
The question one has to ask is how soon any of Data Domain’s competitors will close this gap that Data Domain now has in supporting replication. While others are working on it, the best fan-in ratio of any other solution is only about 20:1 and then fan-in ratios fall down to the low teens or high single digits.If anything, it looks like replication fan-in ratios and the ability to support cascading may supercede deduplication in terms of what users are willing to pay a premium for since nearly every vendor now offers deduplication but replication on the scale that Data Domain offers is still a relative rarity.
Those are my thoughts and observations for this week. Next week I will be writing from Nashville where I will be attending and speaking at the Bare Metal Data Conference. Have a good weekend and I’ll talk to you next Friday!