Slootman Responds to SearchDataBackup.com Interview; Post Process Deduplication Still Perceived as Safer Option by Many
Perhaps the biggest industry buzz coming out of the October 2009 SNW show was not any product announcement or new technology but an interview with EMC's Frank Slootman that appeared on SearchDataBackup.com. Minimally this interview made a number of revelations about EMC's current strategy and future direction for its Data Backup Division. But of greater concern for those enterprises planning to use EMC's products, it revealed a lack of understanding on Slootman's part in terms of what enterprise organizations are looking for in disk-based backup and deduplication solutions.
Slootman's interview was one that I had to reread a few times just to make sure I got it right as the statements he made about the products in EMC's Data Backup Division opened up more than a few eyes (and mouths) in the industry. So prior to responding to anything, I wanted to get Slootman's reaction to this interview to verify he was accurately quoted and this was a response I received from EMC/Data Domain via email:
"I was not accurately quoted in every instance of the write up; some of my statements were not nearly as colorful as they appeared in print. That said, most of the positions represented in the write up are substantially mine."
So with this disclaimer in mind, the interview seemed to imply that Avamar, not NetWorker, is the future of enterprise backup software at EMC. He also stated in the article that EMC is swapping out existing deduplication solutions it previously sold with its newly acquired Data Domain boxes at zero revenue.
Neither of those statements is likely to sit too well with current enterprise customers. It appears Slootman did not get the memo that the way the game works in enterprise accounts is that you let your enterprise customers know ahead of time under NDA what your new strategy is before you go public with it.
The last thing enterprise organizations want (or any customer wants as far as that goes), is to read about strategic changes in product roadmaps from their provider on SearchDataBackup.com. Enterprise customers prefer to hear about changes in direction from their provider first and then have them confirmed in the press later on. It appears Slootman got it backwards in this case.
But this faux pas aside, more disconcerting are the statements that he makes about using post-process VTLs to deduplicate data. He implies that VTLs are not well-suited for deduplication and then goes on to say, "No post-process junk. Post-process deduplication takes all day to land the data for backups. I'm on the beach having a margarita before they land the data on disk."
These statements reveal a lack of understanding on his part as to how post-process deduplication algorithms work as well as customer perceptions about inline versus post-process technology.
First, post-process algorithms do not take all day to land the data on disk for backups as he alleges. In fact, chances are good that a post-process algorithm may land data on disk before an inline process does. Post-process algorithms do not seek to compress or deduplicate data in any way until after the data is first stored to disk and some vendors take advantage of the post-process method to keep an undeduplicated volume ready for faster restores.
Second, not every customer is comfortable doing deduplication in-line, especially those with enterprise FC SANs. Enterprise customers in these environments tend to be conservative and want to understand every step of the process and how any changes, no matter how minor, will impact their production workloads.
In the debate over using in-line versus post-process, post-process is generally simpler to understand and tends to be viewed more positively by many risk-averse customers that I speak with. Their initial objectives when introducing disk-based backup (deduplicating or otherwise) are to ensure their daily backups complete successfully and on time every day. So while I regularly hear that post-process approaches allow them to accomplish those goals, I still from time to time receive reports that in-line cannot.
In-line deduplication is making some significant strides and providers of post-process deduplication VTLs such as SEPATON need to be on their guard to make sure they match the technology innovations that companies like EMC-Data Domain are making. However to suggest that "post-process is junk" fails to take into account the many enterprise customers see post-process VTLs as a safer and often faster first step into disk-based backup and deduplication than what in-line deduplication might deliver.
Slootman's interview was one that I had to reread a few times just to make sure I got it right as the statements he made about the products in EMC's Data Backup Division opened up more than a few eyes (and mouths) in the industry. So prior to responding to anything, I wanted to get Slootman's reaction to this interview to verify he was accurately quoted and this was a response I received from EMC/Data Domain via email:
"I was not accurately quoted in every instance of the write up; some of my statements were not nearly as colorful as they appeared in print. That said, most of the positions represented in the write up are substantially mine."
So with this disclaimer in mind, the interview seemed to imply that Avamar, not NetWorker, is the future of enterprise backup software at EMC. He also stated in the article that EMC is swapping out existing deduplication solutions it previously sold with its newly acquired Data Domain boxes at zero revenue.
Neither of those statements is likely to sit too well with current enterprise customers. It appears Slootman did not get the memo that the way the game works in enterprise accounts is that you let your enterprise customers know ahead of time under NDA what your new strategy is before you go public with it.
The last thing enterprise organizations want (or any customer wants as far as that goes), is to read about strategic changes in product roadmaps from their provider on SearchDataBackup.com. Enterprise customers prefer to hear about changes in direction from their provider first and then have them confirmed in the press later on. It appears Slootman got it backwards in this case.
But this faux pas aside, more disconcerting are the statements that he makes about using post-process VTLs to deduplicate data. He implies that VTLs are not well-suited for deduplication and then goes on to say, "No post-process junk. Post-process deduplication takes all day to land the data for backups. I'm on the beach having a margarita before they land the data on disk."
These statements reveal a lack of understanding on his part as to how post-process deduplication algorithms work as well as customer perceptions about inline versus post-process technology.
First, post-process algorithms do not take all day to land the data on disk for backups as he alleges. In fact, chances are good that a post-process algorithm may land data on disk before an inline process does. Post-process algorithms do not seek to compress or deduplicate data in any way until after the data is first stored to disk and some vendors take advantage of the post-process method to keep an undeduplicated volume ready for faster restores.
Second, not every customer is comfortable doing deduplication in-line, especially those with enterprise FC SANs. Enterprise customers in these environments tend to be conservative and want to understand every step of the process and how any changes, no matter how minor, will impact their production workloads.
In the debate over using in-line versus post-process, post-process is generally simpler to understand and tends to be viewed more positively by many risk-averse customers that I speak with. Their initial objectives when introducing disk-based backup (deduplicating or otherwise) are to ensure their daily backups complete successfully and on time every day. So while I regularly hear that post-process approaches allow them to accomplish those goals, I still from time to time receive reports that in-line cannot.
In-line deduplication is making some significant strides and providers of post-process deduplication VTLs such as SEPATON need to be on their guard to make sure they match the technology innovations that companies like EMC-Data Domain are making. However to suggest that "post-process is junk" fails to take into account the many enterprise customers see post-process VTLs as a safer and often faster first step into disk-based backup and deduplication than what in-line deduplication might deliver.
Leave a comment