Did IBM Do Due Diligence Before Making Diligent Acquisition a Done Deal?

by Administrator on April 18, 2008

I know everyone is hot on de-duplication these days, but I have to wonder about this acquisition.  Is IBM just keeping up with the buzz or does Diligent deliver real value to the consumer?  I will be on the con call at 2:30 EST to see what their thinking is and why this call is being handled by IBM’s Tape and Archive Storage Group boss.

Z

Z

Okay, so the analyst call just finished and here are some of my key takeaways. 

The Need: 

Customers are asking for de-dupe.  InfoPro put it at Number 1 storage thing in November 2007, followed by VTL in the Number 2 slot.  Diligent provides both.

Second, IBM says that customers don’t want to build their own strategy, so they are looking to IBM to provide “pre-integrated solutions” for de-duping data on its way to VTLs or to archive repositories where these are established on spinning rust.

IBM says the key driver for de-dupe is data growth, most of which is driven by compliance-related retention and archiving.  Consumers can use de-dupe to squeeze more data into the junk drawer without adding more capacity, energy cost, etc.

IBM wants to leverage Diligent in three areas: retention (tiered storage with policy driven data movement to archive), compliance (same as previous), ”availability” (same as before but with some VTL sauce trickled in).

Why Diligent? 

IBM says it did lots of due diligence, over a year evaluating other companies.  “Best match” since DT is innovative, ”in-line” (e.g., not a two stage process of de-dupe then ingest), extensible (base product does 450 mbps in a single node, but clustered appliances in the works that can double speeds and feeds; up to 1 PB of storage supported per node)

Also, Diligent is already in 200 of Fortune 500, so the assumption is that they have the attention of the big leagues — and perhaps a golden rolodex.

IBM wants to add value to virtual tape products.  Maybe a strategy for isolating other players (HDS and Sun) though everyone said that HDS relationship will be maintained (“With respect to Sun, it remains to be seen.”) 

Did I catch a twinge of sarcasm?  “Most accounts using the Sun-vended Diligent solution are also IBM accounts.  Who do you think the customer would want to buy the product from?”  Some backpeddaling later however:  “Sun accounted for 1-2 percent of DT revenues in 2007.  However, this is projected to increase to 15% this year.”  Sounds like a relationship worth keeping.

What about De-Dupe services that are about to be offered in Tivoli Storage Manager?  Greg Tevis of Tivoli Storage Technical Strategy was on the call and said the most intelligent stuff.  He alluded to soul searching within the organization around whether two de-dupe solutions were needed — the capability in TSM and the DT stuff.  He seemed to have a pretty good handle on the fit for de-dupe in a purpose-built, service oriented storage architecture and understood my question about the difference between squeezing more bits into the junk drawer (tactical) versus sorting out the junk drawer (strategic).  I like this guy.

They wouldn’t disclose what they paid.  They corrected press accounts that had EMC investments in the company pegged too high at 1.5%.  They corrected, in advance, the inevitable Hollis observation that EMC tech was again finding its way into IBM.  Doran Kempel went out of his way to establish that the factoring technology that is core to DT products was created after the split with EMC and that current technology bears no relationship to the EMC wares.

Asked of Doran and Moshe (the other ex-EMCer who came aboard with the XIV acquisition) would be sharing an office, the answer was a chuckle:  he’s in Israel, Doran is in the States.

Nuff said.

{ 4 comments… read them below or add one }

gknieriemen April 19, 2008 at 8:31 am

I think there is a dose of FUD here.

450 mbps inline? Perhaps at first, but Dilligent has the same problem that Data Domain has…. as the repository grows (i.e. after two weeks of deduped backups), the longer it takes for the dedupe process.

Companies buy VTL’s to shrink their backup window – reduction of storage consumption is secondary. Why is this so easily glossed over by everyone. IF the VTL solution does not shrink your backup window, what’s the point? This isn’t rocket scince… if you add a process to your backup path and it cannot keep up with the backup, you end up increasing, not decreasing your backup window.

Also… Ask any Tivoli end-user who also has deduplication (anyone’s deduplication) and you’ll find that there is very little deduplication.

So… why did IBM buy Dilligent? Probably for competitive rather than practical reasons. Look at the landscape and look at which storage vendors are pairing up with VTL/Dedupe vendors. Overland Storage and (more importantly) HDS are left without a DeDupe solution unless they OEM from IBM.

But I think all of this is really meaningless. The vendors that adopt a hardware-based (on the chip) deduplication system will be the big winners in the long run because the bottleneck will be virtually eliminated. Most of the players in DeDupe do it with appliances (servers) with software.

Who will the winners be?

NetApp? Maybe.

Or will it be the vendors that OEM Hifn’s deduplication/compression card?

joshuasargent April 20, 2008 at 5:56 pm

Since when do companies buy VTLs to shrink their backup windows? Putting backups on disk has always been mainly about the restore process – particularly for individual file restore requests. VTLs only shrink backup windows for a small percentage of the servers in most shops…usually the smaller, unimportant servers. Most VTLs can’t push much faster than LTO-2 speed for a single stream! If you’re buying a VTL to shrink your backup window, you’re probably buying it for the wrong reason. I admit it can help smaller shops with mostly unstructured data, but c’mon…backup windows can all but be totally eliminated today with other tools. VTLs are most definitely not primarily focused on shrinking the backup window…

Don’t underestimate IBM’s strategy with this technology. I’m willing to bet that the technology they just acquired will become pervasive throughout their storage line (except the N Series, obviously). Think about it…how hard would it be for them to modify the dedup code to run on SVC appliances, or DS8000, or XIV? I totally disagree with the assertion that IBM bought them for “competitive rather than practical reasons.” A technology as important as data deduplication (ALL of my customers are asking for it) is important enough to spend the money to have real integration across your product line instead of relying on various third parties to do it for you.

Could IBM have simply relied on FalconStor to provide dedup on the TS7500? Sure. But how much better would the story be if IBM owned the dedup IP and integrated the SAME tech across the whole line?? The last thing I want as a storage admin is a mish-mash of different solutions to manage. Yes, I want best-of-breed technology….but as Toigo says, I’d MUCH rather have good, simple management capabilities. Nothing would be more miserable than having to learn the ins and outs of how dedup works on my VTL versus my high-end disk versus my storage virtualization appliances versus my grid storage, etc…

How much do you wanna bet they do the same thing with their encryption key management server? I’d wager that within 3 years they’ll be using their tape encryption key management server for their disk products… This is what is so great about the N Series (NetApp) – they have the same technology available across their entire line of gear! Managing the smallest N3300 is exactly the same as managing the largest N7900.

I concede to your point about Tivoli…but Tivoli is a different beast than all other major backup software packages. And Tivoli doesn’t own the market… There is plenty of potential to use deduplication in Symantec (or CA, EMC, etc.) shops. What about BRMS for iSeries shops? Mainframe shops? (Could IBM implement this code on the TS7700??) In the larger scheme of things, saying that dedup doesn’t matter because it doesn’t help most Tivoli shops is like saying IP Telephony doesn’t matter because it doesn’t really reduce costs or add value for most very small businesses.

Here’s hoping that IBM has the right idea and can execute well (and in a timely fashion)!!!

gknieriemen April 21, 2008 at 12:33 pm

“Since when do companies buy VTLs to shrink their backup windows? Putting backups on disk has always been mainly about the restore process – particularly for individual file restore requests. VTLs only shrink backup windows for a small percentage of the servers in most shops…usually the smaller, unimportant servers. Most VTLs can’t push much faster than LTO-2 speed for a single stream! If you’re buying a VTL to shrink your backup window, you’re probably buying it for the wrong reason.”

**

You don’t really understand what VTL’s are used for… I’d suggest talking to some end-users. VTL’s are used in environments where the IT admin’s don’t want to stop using their tape backup software but have trouble meeting backup windows. Faster recovery is also a benefit, but backup windows are a challenge many enterprises face every day. I guess the guys at BAE Systems are just a bunch of chumps to you (http://www.techworld.com/storage/features/index.cfm?featureid=2052). 50% reduction in their backup window. Hmm….

I will easily agree that there are better and faster ways to protect your data than VTL’s… but they also tend to be disruptive (i.e. ripping out backup software and changing their whole data protection scheme) and many CIO’s don’t want the hassle or the potential risk that comes with trying something new. I’m not saying this is the right mindset, it’s just the reality. Vendors (and VARs) need to present multiple options with the pros and cons of each.

One last point… I don’t see how any vendor is going to come to market with deduplication on primary storage with a software appliance approach – it just isn’t fast enough for primary storage. It’s going to have be hardware (chip) based integration to work fast enough.

joshuasargent April 21, 2008 at 6:32 pm

I’m not sure it’s fair to compare a really old (probably LTO-1 drives..the article doesn’t specify), broken tape solution to a brand new VTL. The way it sounded, you could have gotten out a pen and paper and written down all the bits by hand faster than their broken solution which had multiple failures every night.

If you were to do an honest comparison of new technology (LTO-4 or enterprise-class tape drives), you would see that tape can very easily blow away the performance of any VTL, usually for lower cost. I’ve sized backup/recovery environments for countless customers, sold the gear (multiple vendors), installed it, configured it, blah, blah, blah….and I was a backup admin in my previous job. I know precisely what VTLs are used for…but it’s more important to understand where they should be used and where they shouldn’t.

Don’t get me wrong – I LOVE VTLs. They practically sell themselves….just not for the reason you assert. BAE clearly benefitted more from having a RELIABLE solution that didn’t fail every night…this is the likely reason their windows shrunk (multiple jobs weren’t getting held in queue waiting for a free tape drive after 3 other drives failed). A current/working tape solution also would have cut the window in half (or better). Anyway….the point of this discussion isn’t really about VTLs, so I can agree to disagree.

Your last point though, I also don’t necessarily agree with….I can’t say I know enough. But I do know that people said the same thing about storage virtualization (some still say it)…and they turned out to be totally wrong. But maybe dedup would prove to be too much…I have no idea and I’m not saying it wouldn’t be much faster on-chip…I just wouldn’t be the one to say that it won’t be fast “enough” in software. It clearly wouldn’t be fast enough for tier-1 storage, but perhaps tier-3 storage would be appropriate (think archival data) where speed is not a factor….capacity is.

Previous post:

Next post: