More Examples Of Why Server Vendors Just Don't Get Storage

Time for me to polarize the discussion again.

One of my common themes is that server vendors just don't really get this whole storage thing.

Sure, they appreciate the revenue and the margins, but they tend to see storage as a complement to their server business, and not a business in its own right.

And that leads to all sorts of interesting discussions.

Example 1  -- HP Discusses Flash, And Does So Very Badly

I was laughing very hard after I read Beth Pariseau's recent take on enterprise flash drives in SearchStorage.

Now, please keep in mind I think Beth is doing a great job -- she's just reporting on the news.  You couldn't make some of this stuff up if you tried.  And I apologize that I can't comment on her speculation about enterprise flash drives showing up in a CLARiiON.

But the best part -- for me, anyway -- is further on in the article, where she quotes HP commenting on enterprise flash drives. 

"HP isn't likely to jump on the SSD bandwagon soon, at least not in the way that the other enterprise array vendors have. Patrick Eitenbichler, director of marketing of HP Storage Works, said that while SSD has value as a Tier 0 for I/O-intensive applications, some of HP's competitors are getting carried away.

"I've been very entertained by what other vendors have put out there about solid-state drives," Eitenbichler said. "We test devices from all the solid-state vendors in our labs, and we break every one of them. As Tier 1 storage, it wears out very quickly. It's not durable, and the cost benefit is not there."

And then the article goes on to say that HP -- and IBM -- is hell-bent on putting this same "unproven, expensive" technology in its servers.  Some of my customers tell me they've already been NDA'd on this.

First, we all know that enterprise flash drives -- not the consumer-grade stuff -- don't have this problem.  Well, most of us do, anyway ... but I digress.

The interesting part is that HP (and IBM, and maybe others) think that putting this expensive, high-value storage in their servers (rather than their arrays) is the right answer.

I think I've seen this movie before ... and I know how it ends.

I'd like to take you back to 1995.  EMC introduces shared, enterprise storage for mainframe and open systems, the Symmetrix 3000. 

We tell customers that storage is an expensive, shared resource, and should be pooled -- and used where needed. 

Don't put it in your servers -- that's a waste.

Customers agree.  They buy lots and lots of shared, enterprise storage, which eventually evolves into SAN, iSCSI and the whole market we see today.  And, it mostly started with a rather largish array that had 32 SCSI connections and 9GB disk drives.

Now -- I have to wonder -- how are enterprise flash drives that much different?  Why would you bury this expensive -- and valuable -- technology in individual servers, rather than making it a shared resource? 

The only people who would think this is a really great idea would be server vendors, I guess.

Please, let's not get into a detailed discussion about how latency and bandwidth over an internal server bus is "better" than over an external storage connection.  That argument didn't work in 1995, and I don't think it'll work in 2008. 

But I"m sure there will be people who try.

And, as the final coup-de-grace, my colleague at HP goes on to say how HP will get enterprise flash drives from HDS sooner or later -- thankfully, they're working with a company that does storage for a living ...

Must be hard being the storage marketing guy at a server company.

Example 2 -- Data Deduplication Hits Big, Server Vendors Late

One of the big storage stories in 2008 is how data deduplication -- in all its various forms -- has been an incredibly hot topic in the industry.  The interest has built much more quickly than most people would have guessed -- including me!

The vendors who do storage for a living (e.g. EMC, NetApp and a few others) were over this thing like white on rice as soon as we saw the market move.  As a result, at EMC, we now have a pretty comprehensive portfolio of 3D (data de duplication) offerings, with more coming. 

Why?  Because we get it -- or, at least, we try to.

IBM?  HP?  Sun?  Dell?  Sure, they've got a thing or two to keep them in the game.  But to say that they spotted the peak in customer interest, and hustled to get there -- well, that's just not the case. 

Although, I should point out that simply being a storage-focused vendor is not enough -- haven't heard much from HDS lately on this particular topic ...

I'm sure that -- over the next few years -- we'll see more from these guys around 3D.  Why?  Because to stay in the storage business in any reasonable fashion (and complement their servers) they'll need something along these lines.

Customers are demanding it.  But it's not a priority, since they're in the server business.

Example 3 -- Sun Promotes Do-It-Yourself-Storage For The Enterprise

I hate picking on Sun again, but their recent foray into open source storage stacks that transform servers into storage shows a basic misunderstanding about how most people use storage.

Maybe I'm biased, but most customers seem to want their storage stuff to work all the time.  They also seem to want a vendor who can guarantee that it works, has tested that it works, and will fix it immediately if it doesn't.

Sure, there will be people who -- for one reason or another -- are attracted to the idea of assembling their own hardware and storage combinations, and hitting the message boards if there's a problem, but those of us who've been in this industry for a while look at this and shake our heads, and wonder.

OK, I'm Not Playing Fair

I work for a company that does storage for a living, and -- by all measures -- is doing well and gaining share. 

And I'm picking on companies that -- although are arguably larger than EMC -- seem to do this storage stuff as a sideline to their server business.

Not to criticize their business models (or philisophical assumptions), but -- I have to ask -- what's going to be more important in the future, if not now: processing ... or information?

I know which one I'm betting on ...

Comments

[Patrick Eitenbichler]

Chuck, I'm glad you enjoy being a storage-only vendor. The customers I talk to are looking for a trusted partner who can help them deliver predictable SLAs to their endusers vs. just a storage company.

Re: SSDs, I'd like to propose having lunch with you 2 years from now. If you ship more storage capacity in SSDs vs. FC disk by then, I'll pay. Otherwise, you'll pay -- and I get to pick the place :-).

See http://www.blackyard.net/?p=908: "Dave Donatelli, a senior exec at EMC, predicted that high-end flash drives will replace high-end hard drives in 2 years."

Let me clarify what I told SearchStorage: SSDs will be a viable technology for I/O intensive applications -- which is why we'll offer SSDs in our servers and StorageWorks XP arrays later this year. However, HP isn't willing to put customers' data at risk. We want to at least be able to source SSD hardware from two manufacturers to ensure uninterrupted supplies.

In terms of your predictions in the past, I wouldn't push that topic... The huge investments you made in network-based virtualization with Invista went nowhere while HP's external virtualization with our XP arrays is a winner, and the virtualization in our EVAs let's customers manage an EVA in 1/5 the time it takes to manage a CX (see http://h20338.www2.hp.com/ERC/downloads/4AA1-6634ENW.pdf). And -- HP has offered Thin Provisioning in our XP arrays for over a year -- a feature that can also be used with all arrays virtualized behind the XP -- while you just recently (and quietly) started shipping Thin Provisioning. I could go on..............

And... talking about deduplication... HP is the only vendor who listened to customer feedback that one size does not fit all. We offer "accelerated deduplication" in our Virtual Library Systems delivering the performance and scalability needed in datacenters, and leverage HP Labs technology for "dynamic deduplication" in our D2D Backup Systems for smaller environments at half the cost of competitive offerings. See http://h71028.www7.hp.com/services/library/GetPage.aspx?pageid=583627&statusid=0&audienceid=0&ccid=14&langid=121 .

Looking forward to our lunch in 2010!

[Chuck Hollis]

I wouldn't take that bet because I think it misses the point.

It's not about capacity. Tapes of all types have more capacity than disks today; paper has more capacity than tapes, and so on.

That doesn't make paper a more interesting storage media than tape, does it?

The real question? Revenue -- what do people spend their money on?

2 years? Dave D's comment about high-end, high performance FC drives looks about right, if not sooner.

Too soon for enterprise flash to dominate in all categories of disk, but we'll notice the effect by then. And, like tape, it'll never really go away.

Not to share a trade secret here, but many of us believe that --- in a few years -- there'll be two types of storage of interest: enterprise flash, and multi-terabyte drives that are deduplicated and perhaps spun down.

In my mind, it's about 3.5 years for the "crossover point" where there's a healthy debate on new storage purchases (flash vs. disk, much the way disk and tape are debated today), and perhaps a revenue crossover 4-5 years from now.

Want to take that bet?

And I think the smartest thing HP does is source a majority of their storage technology from vendors that "get" storage.

If you'd like to go a few rounds of tit-for-tat on product investments and subsequent results, I'll do that, but I won't think it'll be helpful for your case here.

It's funny -- when vendors get caught flat-footed on an important technological shift, there are three strategies: say it doesn't matter, say it doesn't work, or say we'll have it soon anyway.

You get the prize for saying all three in the same article!

Bottom line? I think the industry and customers don't see HP as a major thought leader in the storage biz -- servers (and printers) are far more important at HP.

[Barry Whyte]

I tend to agree that the high end 15K RPM drive space is likely to be the first to be impacted over the next few years with flash, and that large capacity HDD will remain.

However, trying to compare PCI based flash, or a few local SSDs in a server to our old direct attached (pre SAN) world is not the same argument at all. Is it not true that most enterprise servers still have local HDD for boot etc. There are many reasons to keep some local - especially low latency - storage *in addition* to SAN based tiers. To me, your argument is an attempt to spin some FUD mainly due to the fact that you don't sell any servers to plug anything into. Just my 2c.

[Chuck Hollis]

You have a valid point ... many servers have a boot device (whether disk or eventually flash), and some boot directly from SAN or NAS ... and, in that vein -- sure, we'll see flash in servers. Heck, I'll want one for my PC ...

But every announcement I've seen so far from HP, IBM and Sun has been along the lines of enterprise flash for "I/O acceleration" which -- if we're talking writes -- makes it a stateful storage device.

And then I'd be back to my original argument.

[Gavin McLaughlin]

Chuck, I think you've missed the point again on Sun's Open Storage strategy.

Yes, there's the option for the "DIY Model" as you call it, however Sun are also providing solution stacks with integrated support from a single vendor.

What our customers (and yours incidentally) are telling us is that they want balance between cost, growth and risk but WITHOUT vendor lock in or proprietary systems, something that happened in the server world 7 or 8 years ago.

Customers can and will get the best of both worlds through Open Storage, it's not something that should be belittled as "DIY".

Oh by the way, I think 20+ years in storage (8 of which were at EMC) qualifies me a little to talk on the subject ;-)

[Chuck Hollis]

Sorry if I made you defensive regarding your qualifications, Gavin.

Given your experience, you know there's a difference between "support is available" and "support required by enterprise customers is available".

And, unless Sun is keeping a secret, none of us have been able to find hard evidence of the latter for this particular offering, in the same vein as, say your HDS arrays that you resell.

On a personal note, I've lost my enthusiasm for pendantic discussions around "open", "closed", et. al.

I think they've become empty marketing words that struggling vendors toss around when things aren't going so well.