From a backup perspective, the world is at an inflection point. Today’s requirements for growth, rapid data access and speedy recovery are outstripping the industry’s ability to solve backup challenges. Cobbling together stove-piped point solutions is costly to scale and the rapid adoption of virtualization is further stressing backup windows due to legacy processes designed for serial tape.
As the world moves to disk-based backup it needs an integrated solution. If it doesn’t get one, customers will find it increasingly difficult to justify appliance-based disk premiums and will turn to software-led approaches that scale independently of hardware. Or if software vendors (e.g. Microsoft and Veritas) add file system capabilities that perform de-duplication on primary storage, the lack of an integrated solution from hardware vendors will pressure margins.
In my last blog post I intimated that the only thing EMC will integrate in 2009 with Data Domain is the SEC filings (sarcasm). I suggested that a key opportunity for an integrated Data Domain was exploitation of Avamar, EMC’s source-based de-dupe offering, especially in VMware environments, and NetWorker, the asset EMC acquired from Legato Systems. But I posited that EMC’s intense focus on showing progress to Wall Street with the Data Domain acquisition would stifle Avamar and Legato growth prospects in the near-to-mid term, especially in the important VMware market segment.
I believe Avamar, which was acquired for $165M and grew from a small base to more than $200M in revenue will lose critical momentum during the Data Domain integration, because Data Domain, which does north of $300M in revenue was acquired for much higher stakes– $2B+. Data Domain has a significantly larger dedicated sales force than EMC’s Avamar ‘overlay’ sales reps and I believe that Frank Slootman, the new leader of the backup group at EMC will see a faster return on investments in Data Domain. I see this as a negative for EMC’s VMware customers for two reasons: 1) Source-side de-duplication technology like Avamar’s is advantageous to users in VMware environments because backup is a virtualization bottleneck and source-based de-duplication reduces I/O traffic before it hits the physical devices; and 2) A Data Domain target won’t do anything to reduce I/O traffic– and while it may have a place in VMware environments, users should be careful about architecting backup around a de-duplication appliance that plugs into existing backup procedures. My suggestion to users is carefully weigh the impact backup will have on physical to virtual (P->V) consolidation ratios and consider re-architecting backup processes to ensure these targets are met, so you can achieve ROI/TCO goals.
What I learned at VMworld
At VMworld earlier this month I went on a bit of a mission to better understand this issue. First, I found that Data Domain views Avamar as software and its products as infrastructure. In reality, Avamar (in my view) is infrastructure too because it includes disk in the solution but it’s nonetheless an interesting marketing twist on the old source vs. target de-dupe debate.
Here’s what else I found:
- I talked to several customers using Data Domain de-dupe solutions that were achieving high de-dupe ratios and were relatively happy (or very happy). But there were some caveats that I’ll cover later.
- Despite the software/infrastructure positioning, Data Domain concedes nothing to Avamar in VMware environments. It emphasizes its solution in VMware as a highly scalable offering with better tape integration, superior restore efficiencies and consequently better RTO capabilities than other offerings, including Avamar—i.e. on restore, think Data Domain big firehose/Avamar skinny straw.
In reality I think that Data Domain is the fire hydrant and the network is the hose—where restores are only as fast as the weakest network link — but let’s explore this in another post, including the fact that an Avamar Grid can be designed to pump data like a hydrant too.
- In a VMware environment, Data Domain positions itself along with VMware Consolidated Backup (VCB), a VMware feature that allows a more centralized approach to off-ESX-host backup, using a Windows proxy server that accesses ESX data independently.
- In cases where IT doesn’t want to use VCB due to 1) the complexity of integration 2) extra storage required or 3) adding more proxy servers; Data Domain positions vRanger Pro from Vizioncore, an Avamar competitor and a solution with which Data Domain has done some good integration work.
- I also found this Data Domain white paper (registration warning) that describes how to use VMDK/Snap Copy images to Data Domain from ESX.
VMworld hammered home for me that VMware backup is a major problem for customers with many complicated choices. While the Data Domain customers I spoke with were seeing high de-dupe ratios (consistently in excess of 10X), they cited two concerns: 1) The scripting required to make VCB’s work effectively was labor intensive and 2) The proxy server resources were a somewhat unplanned expense item that lowered ROI. Generally these concerns are offset by very high and predictable de-dupe ratios (frequently cited amongst Data Domain customers) and the fact that VMware’s Vsphere enables the virtualization of these proxy servers, which minimizes that expense.
Back to our Main Story
So where does all this leave us? It’s clear that Data Domain, as an independent company recognized the importance of VMware and was moving forward with a strategy to logically extend its reach to the virtual data center by integrating with ISV’s like Vizioncore, Symantec and others. But as an integrated company, EMC has some harder choices to make. From where I sit it boils down to two possible strategies:
1. Let the separate Avamar, Data Domain and Networker product lines compete individually and may the best solution win; or
2. Do the hard integration between Avamar, Data Domain and NetWorker and leverage the Legato asset to create a simple, packaged set of appliances that scales from SMB to the enterprise and beyond.
The former scenario gives Data Domain the latitude to continue building its ecosystem and working with key partners that are competitors of EMC, specifically Symantec, and more deeply integrate with technologies like OST from Symantec’s NetBackup offering. The question is, what’s best for EMC and its customers? The latter option creates the most complete data protection solution in the marketplace and could help EMC gain share against Symantec, IBM and any other target-based de-dupe solution.
But the likely scenario is #1. Frank Slootman, Data Domain’s CEO has been put in charge of EMC’s backup group (BRS – Backup Recovery & Systems) and Data Domain sales and marketing executives are calling the shots on the sales motions. I’m not convinced that scenario #1 is the wrong short-term strategy as integration will be challenging technically and a more open ecosystem (#1) could mean higher volumes for EMC. In addition, the integration work that Data Domain has done with Symantec and others to enable source-based de-dupe capabilities, creates lock-in which is consistent with EMC’s MO. As well, the politics of EMC never allowed, for example, the integration of Avamar and Legato and while taking a more strategic view here might provide better long term revenues and margin, it doesn’t seem to be EMC’s way with its larger acquisitions.
So my prediction is #1 and customers will have to 1) squint through the marketing angles to figure out what’s really best for their environment and 2) do the expensive integration themselves.
Of course they could always ring up EMC’s highly skilled services organization to help them out. 😉