Avamar on a Roll

Avamar has done exceptionally well over the last year. The product has definitely come along, and happens to be hitting its stride at the same time that customers are realizing how problematic certain types of backup are: VMware in particular. All this has contributed to massive growth, and a sudden realization that source based deduplication is a powerful alternative to target based deduplication. It turns out that we at EMC aren't the only ones who feel this way. From a recent analyst report on Data Domain, we have the following:

"analysts see a problem looming on Data Domain's horizon: competition. Since data protection software is a highly saturated market, Slootman says the players are in a "ground war" over space in the data center. And EMC's (EMC) purchase of deduplication startup Avamar Technologies two years ago brought one of the data center's behemoths into the picture... Data Domain is still the dominant player in the field, and the company has kept innovating to ensure a superior offering."

So lets take a look at this. The analyst is acknowledging that Avamar is competing effectively against DD. And we are. Avamar works, and works very well. What is more, Avamar offers a couple of compelling values that DD can't: significantly reduced I/O and CPU utilization on the backup host. DD does nothing for either--you still use the same amount of CPU, the same amount of network bandwidth, and you are still tied to the mediocre capabilities of a traditional backup application in the VMware world.

But the analyst also manages to get a couple things wrong. For one, DD doesn't really do any innovation. Not anymore. Their system is a one-trick pony in the truest sense. It deduplicates data. That's it. It doesn't actually do backup. It doesn't really do archive (they claim it does, but that is, at best, a dreadful case of square peg round hole). It hasn't offered a single new feature of any significance in years. At best they have managed to leverage CPU development to increase throughput and, to a lesser extent, capacity. But it is still a single purpose, in-line only, processor bound, inflexible appliance.

There hasn't been anything new to see from DD in three or four years if you don't count speed bumps. To claim otherwise is to claim that a faster x86 system running Windows Vista makes for a new Vista. And not even Microsoft would say that!

The other critical issue that the analyst are slightly off on is the size and "dominance" of DD. To be candid, DD is simply not the dominant player in the field any more. Not by revenue, at any rate. Nor, if we count net new unit sales (not the install base) would we reasonably conclude that they are dominant. We at EMC have good reason to believe that on the basis of both of those metrics we are competing, and competing very well, with our Disk Library product line. Add Avamar to the mix, and the whole thing tips pretty dramatically in our favor. Consider all other competitors (who we occasionally, regrettably, lose to--but don't worry, that won't go on forever!) and it becomes clear that DD is not only not dominant, but not even exceptional or remarkable: just one of many.

Instead it is emerging that Avamar is the key player in source deduplication solutions. Notwithstanding that changes in architecture on competing products mean that they are not really source deduplication, Avamar has seen tremendous growth, and tremendous innovation. And there is much more to come in 2009. Stay tuned to this blog--when I can tell you what is going on, I will!

Comments

[Denwin]

It has been my view, after testing solutions from both companies, that avamar and data domain are not even in the same market. avamar is not a good fit for a medium sized data center. It requires way too many resources. Avamar at in the near term is strictly a remote site, small data center fit. Data domain excells in larger environments (data centers w/ replication) who are not going to replace their backup software. After testing all EMC, IBM, Symantec, and Data Domain solutions, I would say DD has nothing to worry about in the short term. It really isn't close in performance.

[Scott Waterhouse]

You are absolutely correct in many respects: they are not in the same market to a great extent. And that is part of my observation about DD--they are a one trick pony. That pony may be really darn good at that one trick, but that is it.

As far as Avamar in the data centre, I agree and disagree. For a few reasons, I wouldn't necessarily be rushing out to replace by NW or NBU app for a backup of a 5 TB Oracle/SAP database with an Avamar backup. :) But in the case of VMware, I think it is different. The pains of resource contention and VCBs (being the two dominant method of VMware backup) are sufficient that many folks are looking to replace their traditional backup (and target dedup by extension) with Avamar's source dedup in this one specific instance.

And I won't for an instant concede a performance advantage to DD on target solutions, but that is fodder for another post and another conversation entirely. :)

[Timothy Mann]

Looks like "Avamar Boy" forgot to include the REST of the story - http://biz.yahoo.com/ibd/090128/newamer.html?.v=1

Data Domain CREATED the Deduplication market.

Avamar is a bolt on (usually free, that's how good it is) to the "more disk will fix it" strategy.

That is some innovation....

[Scott Waterhouse]

Let me attempt to respond to the substance of the comments Timothy.

First, Avamar is not an "bolt on" in any way shape or form. It is a complete, stand alone product.

Second, Avamar has been around longer than DD. Ooops. That is right. The company that "created" the deduplication market is newer than Avamar. Avamar was founded in 1999 as a backup product that used deduplication to reduce the amount of capacity required to hold your backups. Data Domain was founded in 2001 as a target that used deduplication to reduce the amount of capacity required to hold your backups.

So, who is the innovate company? Who is the creator?

[JimTing]

Denwin has it right. Even for VMware, the DD solution tested much better than Avamar or Puredisk. VCB removes any contention from the production servers. Its just a snap to the proxy. Any contention is on the proxy which has no affect on the production VMs.

I also don't think a source based dedupe solution is a good fit for large datacenters and specifically applications that need a dedicated phyiscal sever like databases. The source dedupe was very slow and created resource issues. So that begs the question, would you want TWO backup software solutions in your enterprise? One for VM and one for everything else?

Besides, if Avamar works so well why is EMC pushing the Quantum based 3DL so much?

----------------

So typepad is being weird and not letting me post comments, so I will just add my thoughts to the end of this one:

Jim... It would be interesting to understand more of the details around your test. I have been pretty open with Avamar plusses and minuses elsewhere on this blog, and pretty clear on Avamar performance characteristics. Having said that, in many/most scenarios it is reasonable to expect that Avamar will perform better than any DD box, including the 690.

With respect to VCBs and proxies, you are right, of course. But extend the logic further: moving the load to a proxy server merely moves the load somewhere else. If you do traditional backup from the proxy server, then you have to ask how many ESX servers you can back up from a single proxy. And with Avamar, that number tends to be about double what a normal backup application can manage. That is--if you are backing up 4 ESX hosts with NBU and VCBs, you might managed 8 ESX hosts with Avamar.

Other advantages over DD include an inherent HA ability too. DD has single points of failure, Avamar server does not.

Also, you will get few arguments from any informed Avamar expert that it is not appropriate for all situations. As I commented above, if you have a 5 TB Oracle or Exchange server, you will likely want to stick with your traditional backup app.

Finally, you ask "would you want two backup... solutions" and the answer my customers are giving me is emphatically yes. Believe me, I am as shocked as you are. :) I have been doing backup for 15+ years, and never before have I seen such willingness to run to backup apps in the data centre at the same time.

PS: Quantum/EDL vs. Avamar is apples and oranges. We are all pretty clear that Avamar is for ROBO/VMware/and file services--where you are willing to consider a different backup application. EDL is for large environments and where and when you don't want to think about changing your backup application.

Scott