Data Domain: Accelerating The Transition From Tape To Disk

As part of EMC's recent flurry of product announcements, EMC's BRS division (backup and recovery systems) is especially well represented: not only by bigger/better/faster versions of their existing successful products, but new offerings that open up interesting market segments.

As the IT industry progressively transitions to using disk for backup, recovery and archiving (and, correspondingly, moving away from tape), customers need solutionsEMC's BRS division is in an enviable position in terms of differentiated product features, leading market share and great customer enthusiasm.

Right now, I think this game belongs to EMC.

From Tape To Disk

If you've seen the market revenue stats for tape drives, libraries, media, etc. -- it looks sort of like a ski slope.  

The market demand for these products (tape automation is shown here) appears to be contracting in the double-digit percentage range annually.

This is not to wildly state that "tape is dead" or anything similar: it's just that we're all  apparently using a lot less of it with every passing year.  

The numbers don't lie.

Reactions tend to split in the community: some argue passionately for tape's continued longevity in the industry, others just want to move on to the next thing as quickly as possible.

From an industry perspective, I would state the obvious: declining category revenues for tape means that progressively less R+D will be invested in advancing the technology, with the inevitable outcome being that tape eventually becomes an interesting niche in the way that so many other storage technologies have become.

BTW, I have a nice supply of 3.5" diskettes if anyone is interested :-)

The motivation for the transition boils down to a few key elements.  

First and foremost has been rapidly declining disk media prices.  A terabyte of raw capacity isn't the big deal it used to be.  The advent of powerful software that not only dramatically compresses backup images, but retains compatibility with existing IT processes has made cheap disks all the more usable.  

And, finally, the ever-increasing performance of Intel-based CPUs has raised performance to impressive levels.

All three trends are at play here with the new product announcements.  

So, What's Being Announced?

Three things: two that are somewhat expected, and one that has the potential to be a big game-changer in the marketplace over time.

The first part of the announcement is the expected bigger/faster/better versions of the popular Data Domain backup engines.  

Speed matters in backup devices, especially those that are shared resources for multiple applications.  

Faster backup speeds means both more application availability and more efficient consolidation.  It's a win-win.

As data volumes continually grow, there's perpetual built-in demand for bigger -- and faster -- backup devices.

At the high-end, the new version of the Data Domain Global Deduplication Array (or DD GDA for short) now boasts an ingest rate of 26.3 terabytes per hour, and a maximum of 570 usable TBs that can result in a logical capacity anywhere between ~5.7 and ~28.5 petabytes of logical capacity.  

This is roughly twice the performance of the previous version of the GDA, and an estimated 7 times faster than the corresponding IBM offering -- our next closest market competitor.

The more traditional DD units have been upgraded as well: the new DD890 and DD860 boast not only sizable performance and capacity bumps, but are both data-in-place upgrades for owners of the previous DD690 and DD880.

For those of you that missed it, "DD Boost" refers to a nifty software option that uses the frequently untapped processing cycles of backup servers to speed performance even further.  As a result, throughput numbers are shown both ways -- with and without DD Boost.

Unlike primary storage devices, there seems to be a bottomless demand for bigger/faster/better versions of backup-oriented dedupe devices.   Fortunately, thanks to the Intel roadmap, there will likely be many of these healthy bumps in store for years to come.

The second part of the BRS announcement has to do with host support: now adding IBM's inimitable "IBM i" (aka System i / iSeries / AS400 -- no jokes about the artist formerly known as Prince, please).  

For those of you not familiar with this particular server market, there are a *lot* of iSeries out there in very specific industries and geographies.  And, yes, they need better backup solutions as well :-)

Since IBM markets this platform as an all-in-one "solution", there aren't a lot of third-party infrastructure options for these users.  IBM implements some interesting nuances in the environment that require special storage engineering support.

EMC is the only major vendor (other than IBM) that has consistently invested in providing primary storage support for iSeries (through the Symmetrix product family) and now disk-oriented backup via Data Domain.  

For those of you interesting in mainframe (formally zSeries) backup solutions, I'd invite you to keep an eye on our recent acquisition of BusTech.

And The Big One ....

The third part of the announcement may seem rather unremarkable at first, but -- at least in my opinion -- has the greatest potential to change the dynamics of the storage market over time.

The new product is simply named "Data Domain Archiver".  

From a hardware perspective, it appears rather ordinary -- essentially a cost-and-performance optimized version of other Data Domain products.  

Simple-to-use software establishes straightforward migration policies to move aged backup data down to more cost-effective tiers over time.  And it all appears as a logical extension of the backup environment people are using today.

But there's far more to the story than meets the eye -- and here's why.

Most businesses need to keep data around for a long time.  They may never need to access it -- but when they do, it has to be there -- and be there reliably and relatively quickly.

Historically, tape has been the preferred medium for many of these long-term hope-I-never-have-to-use-it archives.  Internally, we refer to this particular use case as "tape's last refuge".

With this announcement, the game has changed -- we can now offer these deep-tape users a brand-new value proposition -- get rid of tape for long-term retention purposes.

How the numbers work out (tape vs. DD) in your particular environment will likely vary.  We're already able to make the case to switch for many customers, including this one.  

However, regardless of how the numbers might look for you today, all available data shows that industry forces are pushing one number down, and the other one up.

The forces pushing cost-to-store downward on disk-based solutions are easy to understand: rapidly declining costs for disk media, as well as more powerful standard processors that can efficiently deduplicate larger and larger volumes of information.

A little harder to get to -- but just as important -- are the forces pushing tape costs upward.  Due to declining revenues -- vendors don't appear to be advancing the economics of tape automation and media as quickly as their disk-based competitors, like Data Domain.

For most people, costs like floor space, labor for tape handling and other related services aren't going down fast enough.

Indeed, I think this "deduplicated disk as long term archiving" aspect is probably worth a bit more discussion

Integrating Long-Term Archiving With Backup

The use case for deep tape archiving is pretty clear: put it on tape, and pray you never have to access it.  If you do, hopefully it's just a very tactical and focused piece of data you need to get back, and -- again, hopefully -- you don't have to do this very often.

But things have a way of changing over time.  

First, the number of events that require a archival tape restore are apparently increasing over time.  It's just assumed that the IT guys can easily go find whatever files might be needed from several years ago.  

Finding and restoring particular files off of long-term tape can be very labor intensive -- if the data can be read at all!  Anecdotally, I've been told this retrieval process is usually an iterative process, e.g. "Is this what you're looking for?"  "Well, is this what you're looking for?"  "How about this?"  ... and so on.

Since everything archived sits on random-access media -- and typically appears as a single, consistent space -- retrieving specific data from the archive is far faster than before -- and takes much less effort.

Measuring The Impact

With this particular announcement, there are actually two important things going on.

The first is rather clear: disk is well on its way to replacing tape for primary backup and restore.  

EMC's BRS division is clearly leading the way here, and -- with this announcement -- not only have greatly magnified their fundamental economic proposition for their customers, but increased the distance between themselves and their next closest competitor.

But, more interestingly, with the Data Domain Archiver, EMC has essentially opened up an interesting new incremental new market built around displacing tape's last refuge in longer-term deep archives.

So, here's the real question -- how long before the only place we'll see tape is here?

Comments

[Dominic Cody]

Great post as usual, I do agree with the comment about disk replacing tape for primary backup and restore. I do question that EMC BRS is leading the way and is the best option for customers.
I think if a company wants Backup to disk with De-dupe and has the infrastructure for a separate backup LAN on 10GBe then Data Domain ticks the box. However if they want a VTL then I don't think it fits at all and have seen this in a company I worked for were EMC are now trying to change the support of the box due to major issues with it running as a VTL.

Also is the Data Domain Archiver the replacement for the highly thought of Centera platform which is ageing now or is this something to sit alongside that product?

Dominic

P.S - Just so I am open, I did work for a leading retailer in the UK but in November changed and now work for HP, but my opinions are my own and not company driven.

[Chuck Hollis]

Hi Dominic -- thanks for the thoughtful comment.

You're right, there is no one-size-fits-all answer in the world of IT. Specifically, I can imagine more that a few use cases where other approaches would be indicated -- from EMC and other vendors.

Regarding DDA and Centera, the use cases are different. DDA basically extends the disk-based backup into a cost-optimized archive tier. Centera, however, is strong when people actually want to *use* the archived data for a variety of purposes -- search, compliance, customer service, etc.

Customers who just want to replace tape for their deep (and mostly unused) archives will want to consider Data Domain Archiver. Customers who want to archive and use their data for a variety of purposes will probably be more attracted to Centera.

Most people are unaware that Centera continues to do well for us in the marketplace.

Thanks again

-- Chuck

[Dominic Cody]

Hi Chuck,

Thanks for clarifying the Centera/DDA this makes perfect sense to me now.

So DDA for writing something that they need to keep indefinitely but will potentially never ever need or use again and Centera for say mainframe DB2 data that they need to get off for reference only but will never change it and is the only thing left on their mainframe costing them lots of cash to keep it on there.

I agree on not one size fits all and I probably should have mentioned in my first post that if a customer does need pure VTL then of course you still have the EDL kit at EMC to use for that purpose should someone be adamamant they want VTL and don't care about de-dupe. (look at that a HP employee recommending other EMC kit, dam must be in the wrong job)

Thanks

Dominic