Actual Unitrends and REDACTED VMware vSphere 5.5 Storage Deduplication for a Single VM and a Single VMDK
[Preface: Note that the although the data is real, the name of our competitor has been obscured, changed, and otherwise redacted to avoid hurting anyone’s feelings.]
Last week there was a “twitter storm” around a marketing campaign that Unitrends which called out another company (a proverbial 800 pound gorilla in the 100% virtualization backup market) on its lack of efficacy regarding deduplication. I was flying back from Toronto where I was meeting with some prospective partners; so I got into the tail-end of it all. Beyond the typical Twitter kind of stuff (e.g., an allegation of aggressive marketing to which I sent a link of this competitor doing the same thing on their web site, an assertion that the competitor was superior to Unitrends because it has more customers to which I asked if that meant that McDonald’s was the best restaurant in the world – that kind of thing), there was a more substantive discussion.
One of the things that came across in all the conversation was how some folks found it to be distasteful to call out competitors by name regarding their weaknesses. I disagreed – but understood that it was a debatable point even though this particular competitor seems to do it quite often. Then over the weekend someone forwarded me a forum post from the competitor’s web site responding to this kerfuffle – and engaging in some name calling – that sort of thing.
I understand the perspective, if not the facts, of the competitor’s forum post. Being limited to virtualization HOS-level backup can at times limit ones perspective – it’s like the old saying about all the world looking like a nail to the person who only has a hammer. At the same time, I think that backup can be complex – and what the buyer needs these days is as broad a perspective as possible about their choices.
The chart above shows real-world storage footprints, and thus implicitly deduplication ratios, over time for a single VM, with a single VMDK, with a 5% change rate. This isn’t theoretical – it’s the actual results from the simplest possible test I could conceive of deduplication. In my next post on this subject, I’ll explain why you see this type of result and the trade-offs that occur not only with respect to deduplication, but with respect to HOS-level backup versus GOS-level backup (note: HOS-level backup is generally faster, GOS-level backup generally has higher deduplication ratios in a content-aware deduplication scheme, etc.)
I decided to publish this because over at our competitor’s forums there’s a statement that I find – well, the best way I can describe it charitably is as incredibly ignorant – because otherwise it’s just plain not truthful. The quote was in effect that “there is no difference between global and per-job deduplication.”
Given the enormous marketing budget of our competitor as well as their demonstrated world-wide success at driving marketing messaging (it’s almost, but not in any way, as if their name rhymes with “meme”), I couldn’t help but be reminded of the Winston Churchill quote: “A lie gets halfway around the world before the truth has a chance to put its pants on.” I think it’s time we put the marketing aside, and start talking facts.
Thus, for the next few weeks I’m going to write a series of posts about deduplication. I think it’s important to help the truth get its pants on more quickly. As always, look forward to any feedback, comments, issues, and concerns.