the storage anarchist

1.014: the laurel and hardy of thin provisioning

UPDATED (July 3, 2008): Deletions struck-out, additions noted in green.

One is decidedly skinny.

The other is unabashedly portly.

And I'm not talking about Stan and Ollie, folks.

No, such is the differentiation between the thin provisioning implementations of IBM's SVC and Hitachi's USP-V/USP-VM.

Sir Barry White eloquently describes the petite implementation of SVC's fine-grained Space-Efficient Virtual Disk (SEV for short) in a recent blog post (any resemblance of BarryW to fellow Brit Stan Laurel is purely coincidental, I'm sure).

Not to be outdone (and in an obvious attempt to justify the Hardy-ness of Hitachi's Dynamic Provisioning), HHSNBN explains why DP's heavyweight approach makes for better thinness (at least on the USP-V). Given the title of his post (When is Thin Provisioning Too Thin?), I figure ole' HHSNBN doesn't think the SVC's implementation is all that, shall we say, robust.

IMHO, both have managed to gloss over details that are very pertinent to understanding if, when and where one implementation is better than the other. Not surprising, especially since BarryW & I both know full well HHSNBN will never respond directly to any inquiries or challenges. No, HHSNB prefers only one-sided discourse (his side, of course), so I guess that leaves it up to me to try to tease out the truth.

So let's look a little deeper at these near-opposite implementations and see what we can figure out for ourselves, shall we?

Warning: readers of this blog have asked that I spend more time talking tech,
and less time bashing the competition.

This post is about as close as I can get to fulfilling those requests...

 

1.005.2 hitachi hits new lows (reposted)

Notice: I withdrew this post yesterday afternoon after receiving a challenge from a commenter who insisted that I was mixing up Hitachi's disk drive results with their storage array business. After carefully reviewing Hitachi's published results, I am convinced that the revenue growth numbers I used for Hitachi's storage business are correct (and exclude HGST). Therefore, I have reinstating my post intact, with the addition of a new "Hitachi Math" section (in blue) below.

Hitachi announced their earnings this past Tuesday (May 13th), and their Storage Solutions results (among others) were particularly gloomy. Not as bad as the free-fall in plasma TV sales in the US that they experienced, maybe, but dismal nonetheless.

In what is their fiscal Q4, once high-flying Hitachi only managed to eek out storage revenues that were down 1% from a year ago and down 3% from last quarter, while both EMC and IBM (if you include Tape) actually grew revenues double digits Y/Y in the same period. It marks a notably downward trend in Hitachi's Storage Solutions revenue growth over the past couple of years, as can be seen in this chart:

And Hitachi's projections for the future was for even more revenue contraction for this quarter and next - shrinking perhaps another 5% before they expect a turn-around, they said.

more hitachi math

As I noted above, the above results were challenged yesterday with an assertion that the "decline" was due to HGST. Fact is, these numbers are taken directly from Hitachi's earnings supplement, where they report "Storage Solutions" (array hardware, software and services) separately from "Hard Disk Drives" (see the top of page 2).

You'll note that Hitachi reports by halves (wouldn't want to make it easy to figure quarterly results now, would you?) - but if you go back to last quarters' results, you can do the math to verify that their Storage Solutions revenues were up 2% Y/Y in FQ3'07 and down 1% Y/Y and 3% Q/Q in FQ4'07 - just like I said.

In researching the accuracy of the numbers used in the chart, I was also allowed to see a couple of financial analyst's reports that included additional revenue insights provided them by HDS executives. While I cannot reprint specifically what these analysts published, I must say that there is something that smells an awful lot like Hitachi Math in their reports. Perhaps they just misunderstood what the HDS execs told them, but the numbers the published in their reports simply don't add up.

More importantly, since HDS only sees revenues excluding Japan, their perspective undoubtedly skews any possible analysis - especially since it's not clear whether HP and/or Sun storage revenues are reported through HDS or if they go directly to Hitachi Ltd. (I'm pretty sure they go directly to Japan).

And if revenues really weren't shrinking, then why the heck would HDS execs be trying to spin the story with Wall Street in the first place?

Bottom line: Hitachi Headquarters reported (and documented) that FQ4'07 Storage Solutions (ex-hard disk) was down 3% Q/Q and down 1% Y/Y - just like I said.

And on top of declining revenues, word on the street is that morale in Hitachi's US field operations is at an all-time low. Nobody seems to know if morale is suffering from the recent out-sourcing of customer service, the collapsing of the former solution/consulting business with the former Hitachi Data Systems subsidiary, or the new Japanese management that are running the new US holding company now. Or maybe it's something else?

Given that the flagship USP-V is nearly a year old, it seems very odd that revenues would be shrinking at a time when the new system should be really starting to gain traction.

It all makes me wonder...
 

0.061: swinging from the sidelines

He Hu Shall Not Be Named has returned from his vacation in Aloha-land to learn that EMC has announced something of which he knows nothing about. Predictably, he has wasted no time demonstrating his total lack of knowledge about enterprise flash drives in his latest blog post.

Makes me wonder what the "CTO of Hitachi Data Systems" really does, since he obviously isn't following emerging new storage technologies all that closely.

Nor did he even take the time to do any research before posting, choosing instead to throw wildly uninformed punches at the technology in an attempt to defend Hitachi being caught flat-footed.

Not surprisingly, HHSNBN even manages to drag virtualization into the discussion. Now, is it just me, or does it seems like he can't write a post without using the V-word? I'm beginning to think he must use a customized version of the Ergonomic Keyboard for Pirates that has been making the rounds of late (pictured at right). On his version, though, the "Avast" key types "Virtualization."

<Shift>VIRTUALIZATION<Space>RRRR<shift>!!!<Enter>

Swinging from the sidelines

Clearly lacking a relevant clue about the the technology (and unable to buy a vowel), HHSNBN alleges that anything NAND can do, DRAM can do better and faster. Which is true, I guess, if you include consuming your IT budget in that comparison. Because as expensive as SLC NAND flash is today, it's already an order of magnitude or two cheaper than high-performance DDR2 SDRAM. Must be that Hitachi Math thing again.

So I'll just tack that one on the bulletin board right alongside "Intermixing, slower, less reliable SATA or FATA disks in tier 1 storage systems will impact that system’s performance and availability", which was HHSNBN's excuse for not supporting SATA in the USP. Right up until Hitachi Japan added SATA support to the USP-V at the end of last year, that is.

Some words of advice seem appropriate here:

It is far better to remain silent and be thought a fool
than to speak out and remove all doubt.

(the debate rages over who actually said this first)

the truth about enterprise flash drives

For the record, the enterprise flash drives that EMC announced last week are not the same as the consumer flash drives that is going into Airbooks, laptops and IBM blade servers. And they won't silently lose data (they'd be pretty useless if they did) - like disk drives they remap suspected bad blocks before they cause a problem. They'll probably outlive the practical life of the storage array before they wear out - the nominal operating life of SLC NAND flash typically far exceeds the rated minimal 100,000 writes per cell, and when you add in wear leveling across nearly 2x extra capacity within the drive, they will outlive all but the pathological 100% write forever use case.

Yes, enterprise flash drives use SDRAM buffers to accelerate writes, with both internal and Symmetrix-supplied backup power to protect that memory against unexpected loss of power, be it momentary or an extended outage. They are shielded from electrical and mechanical interference in a disk-drive form factor. And they do indeed provide extensive status reporting (basically the Fibre Channel & SCSI equivalent of S.M.A.R.T.), affording the array the ability to be proactive in protecting customer data.

And you don't have to take my word on that.

Thankfully, the folks over at Wikibon have taken a much more thoughtful approach at evaluating last week's Symmetrix announcements. After what was obviously a lot of actual research, they today presented a comprehensive review of enterprise-class flash drives, the customer benefits they can deliver and the practical implications that they will likely have on the entire storage industry over the coming years. I encourage you to read and comment on their review.

Oh - it's aptly titled EMC Lands a Haymaker.

0.053: when 2 + 2 isn't 4

IBM seems to be learning Hitachi math.

At least, some of their field sales representatives appear to be experimenting with Hitachi's notorious approach of magically mangling numbers in order to subliminally mislead perceptions.

In a recent RFP (one that I admittedly heard about third hand), the customer-to-be reportedly specified that all bidders must propose only storage systems that are 4Gb/s Fibre Channel front and back.

Now, if you follow high-end storage at all, you know that ONLY the EMC DMX-4 and the Hitachi/HP/Sun USP-V today employ a 4Gb/s FC back-end, while all models of the IBM DS8300 use only a 2Gb FC back-end. A simple fact that's easily verified reading the specification sheets for each product - no room for argument.

So imagine the buyers' surprise to receive a proposal from their local IBM rep, complete with  written confirmation that the proposed DS8300 Turbo did indeed come with a 4Gb FC back-end!

As I've been told, the fun didn't stop there. When the customer confronted his (or her, I'm not sure) friendly IBM-badged sales representative, they were even more surprised to be told that the 8300's back-end indeed was 4Gb/s. The rep explained, with a totally straight face, that each disk drive was serviced by not one, but TWO 2Gb/s Fibre Channel connections, which clearly adds up to the required 4Gb/s.

You just gotta admire the chutzpah!

0.049: hitachi rifs mr. t

Like me, you probably noticed that the warfare and annihilation was markedly toned down in the latest round of press releases out of Hitachi Data Systems. So much so, I had to wonder why?

So I did a little digging, and learned that HDS's PR dude Steve Zivanic was recently RIF'd by the new VP of Marketing over an apparent culture clash. Steve was reportedly the guy responsible for those over-the-top (IMHO) Mr. T YouTube-omericals ( (original, "T" for Trucker and mid-TEEr T).

Or at least he was the guy held responsible for these tasteless productions (again, IMHO). Presumably he was also the brains behind the comic-book expressions littered throughout most of the HDS press releases we've seen over the past several years.

Steve has been reportedly replaced by a fellow named Emory Epperson.

While I personally can't say that I'm sad to see Mr. Zivanic go, I was also heartened to learn (from a source who requested anonymity) that former HDS blogger David Merrill was still under the employment of HDS (just muzzled for his honesty, apparently).

But I was more than a bit concerned that poor old Mr. T might be turned out to pasture as well. 

0.039: ibm and spc vs. hitachi math

Hitachi dropped another shoe Monday with its announcement of the best-ever SPC-1 benchmark results for an Enterprise Storage System.

I'm sure the "thud" was pretty deafening over in Blue-ville, especially given this recent reiteration of IBM superiority by Tony Pearson - partially on the back of the DS8000's (now defunct) claims to the top spot on the SPC stepladder.

I pretty much established my position as a disbeliever in the real-world value of benchmarks such as the SPC in my prior post entitled the case against standardized benchmarking, so I won't rehash my arguments of irrelevance here.

But I will take the opportunity to add to my continuing expose of Hitachi math, both as a service to my readers, and in support my pals Tony & BarryW over at Poor Old Big Blue. I'm sure they'll have their own spin soon, but I figured I might be able to jump-start their responses.

But before you read on, I'm curious: Were YOU able to recognize the Hitachi math in this announcement?
 

0.037: storage anarchism recategorized

At the request and suggestion of several readers, I have re-categorized this blog, classifying my posts along specific storage themes, topics and technologies. Hopefully, this will make it easier for new readers to find topics of interest and follow along. I've also simplified my sidebar a bit to make it easier to find things. Thanks for the suggestions.

And while I'm here, thanks also to Margaret Rouse for making my "Hitachi math" infamous as the September 27, 2007 entry on her Overheard in the Blogosphere blog. The anarchist is honored to share space alongside the likes of Larry Ellison and Bill Gates...and Margaret's blog has claimed the pole position in my own blog reader.

0.035: hitachi drops another shoe (it sounded like a slipper!)

I guess I should be ashamed. Embarrassed at least.

Yesterday, someone asked me why I hadn't yet commented here about Hitachi's baby USP-VM announcement made earlier this week, and I responded "what announcement?"

Yep. I missed it. Completely.

Not that I wasn't aware it was coming...Mr. T doesn't make the rounds on Wall Street on the arms of HDS PR folks for nothing, nor do you wake poor old Claus from his year-long blogging hiatus unless something really important is about to happen.

As I had prior blognosticated, there were (and still are) gaping holes in the initial USPV announcement - clearly the team back in Japan had much more in mind than had been announced back in May. And at least a few rumors indicated that many of these would make it through the QA gauntlet before the end of Q3'07.

Guess this stuff really is hard for everyone.

So I guess because I'd expected so much more, the actual announcement of the USPV "mini-me" slipped right under my radar. I probably shouldn't have taken Nigel's advice and stopped stalking HHSNBN (although it has been therapeutic not to read about yet another world calamity that UVM can solve for a couple of weeks, mind you).

But now that mini-me has been brought to my attention, you just know I have to comment. Interestingly though, this time I'm not alone in correcting the typical misleading marketing that the Hitachi Data Systems Santa Clara marketing miscreants have become known for.

It seems I have some new friends in my quest to keep things honest.
 

0.006: usp-v launched! (yawn)

Just perused the launch materials for Hitachi's new USP-V, and I've got to say that if *I* were a customer, I'd be thinking:

"You're over 6 months late, and THAT'S all you managed to do?"

Something weird must be going on in Hitachi's Japan development center. I mean, the ostensibly did SOMETHING to improve the performance, but it looks like someone forgot to tell the hardware guys to make room for more disk drives. This brandy-new USP-V still can only support a maximum of directly-connected 1152 disk drives - what's up with that? And there's no support for cost- (and energy-) efficient disk drives - the largest drive supported is still only the 300GB, and only the 10K rpm version at that.

The Symmetrix DMX-3 has supported up to 2400 drives for over a year, and today supports both the 500GB 7200rpm LC-FC and 300GB 15K rpm drives for optimal "green" tiering. Maybe Hitachi just can't support drives this large, unless they're in someone else's storage array...but why buy two controllers to get more power-efficient storage?

I dunno, do you?

0.004: get ready for tagma-too!

Word's out - there's soon to be a new USP from our pals at Hitachi/HP/Sun. And I'm sure you didn't hear about it first from me.

It's about time.

Hitachi had been on a pretty set schedule of bringing new high-end platforms to market just about exactly every 24 months. And in fact, it seems many customers were specifically told in early 2006 to expect the next-gen "Tagma" in the September 2006 (ish) time frame, almost exactly 2 years after the original USP12000/9990v was announced.

That would make next week's (rumored expected) announcement just about 8 months late. Makes one wonder what when wrong - what happened in 2006 to cause such a delay? A delay so long that Hitachi's Wall Street Minions were out in force late last summer, admitting to anyone who'd listen that the company would likely suffer significant loss of market share through calendar Q4'06 and Q1'07.

Although the numbers aren't all in yet for Q1'07, the predicted loss of storage market share the last couple of quarters looks very likely to have been correct. In fact, estimates I've seen indicate that IBM's DS8000 surpassed Hitachi's USP+NSC sales in the high-end for the first time back in Q4'06.

After such a long wait, I'm pretty sure what the folks over at Hitachi are thinking right about now...

...Welcome to the machine!