0.068: rotflmao!

And I almost couldn't get back up...

Regular readers know that I've been predicting the market exit of the DS8000 for most of my tenure in the blogsophere. As can be expected, this has attracted the ire of IBM bloggers, who routinely dismiss my observations as misinformed FUD. I've even been called "silly" by members of the IBM blogging corps.

So perhaps you'll understand why I was laughing hysterically after reading comments made by Charlie Andrews, IBM's marketing director of storage systems, in Beth Pariseau's recent SearchStorage article (see IBM: Thin provisioning on the way for DS8000).

According to Charlie, the DS8000 is indeed still being updated, although apparently only to support IBM's newest mainframe monopoly models. Some neat stuff - like larger volumes and granular fail-over (welcome to the 21st century, zSeries!).

But when I read Charlie's responses to Beth's inquiries about the increasingly long list of features missing from the DS8000, I literally fell out of my chair laughing.

Not almost - I actually hit the floor. LMAO

Read on to learn why...

[UPDATED 7AM EST February 29, 2008 - see below]
 

mi cliente no es su cliente

Apparently, Charlie markets the DS8000 to a different set of customers than the enterprise customers served by EMC's DMX (and Hitachi's USP-V, to be fair).

At least, that's the only explanation I can come up with for his responses to Beth's questions regarding the lack of variable partitions, thin provisioning and solid state storage on the DS8000:

"We need to do more," Andrews said. "The challenge for us has been the breadth of our portfolio, and where we've chosen to allocate resources."

"There are always lots of things you can work on and some of them will become compelling technologies right away, and others won't for at least two years. The important thing to us is not to be ahead of EMC and HDS, but to be ahead of our customers and what they're moving toward as a group."

If that isn't an outright admission that IBM has turned it's back on the enterprise storage market, I don't know what is. I mean, if you're not even in the race, how can you expect to win, place or even show?

Honestly, over the past year, I can't recall a single one of the literally hundreds of DMX customers and prospects that I've spoken with who WASN'T interested in thin provisioning. And as the person responsible for the vast majority of pre-announcement NDA briefings on EMC's pioneering push to bring flash storage to the enterprise, I can assure you that there is not only interest, but significant pent-up demand for faster storage solutions that are tightly integrated with the feature set of enterprise-class storage arrays.

I guess ole' Charlie must not be targeting the same group of paying customers as I have been (which is fine with me as an EMC shareholder, now that I think of it).

But what the heck? Even Wall Street and the press has recognized the need for thin provisioning, chastising EMC frequently over the past 9 months for being "late" delivering it on the DMX (still scheduled for GA this quarter).

All I can say is, if EMC was late to the party, then IBM must not have even been invited.

Andrews did admit IBM has seen a demand for thin provisioning from customers. "I think this year is when we need it," he said.

Ya think?

 

But while I got a chuckle out of those quotes, I was still seated safely in my chair until I read the next section of Beth's article...

ibm don't need no stinkin' flash drives

OK, let me first say that when a vendor catches its competition flat-footed, no matter what industry we're talking, there are only a couple of responses the embarrassed laggards can use:

1. "<New feature> isn't ready/proven/viable yet"
2. "We haven't seen any customer demand for <new feature>"
3. "We're all over <new feature>, we'll be announcing it soon"

Generally speaking, you can only use that last one if you actually know you're going to deliver the feature in the not-too-distant future ("soon enough" in marketing-speak). I'll acknowledge that's exactly what EMC did regarding thin provisioning back when Hitachi announced it for the USP-V - go ahead and let everyone know you've been working on it too, just on a different release cycle.

But when a vendor DOESN'T use #3, you can be pretty sure that not only were they caught flat-footed, they hadn't even realized the significance of <new feature> enough to have an educated opinion on it. It's very telling that HDS bloggers HHSNBN and Claus used both #1 and #2 from the outset with regards to EMC's announcement of flash drives.

But IBM must run their marketing folk through misdirection courses as a prerequisite to public speaking, because good ole' Charlie came up with a position I could never have imagined:

IBM doesn't see solid-state drives as the answer to power or cooling, or performance issues in high-end arrays. To drive down cooling costs and free up space on Tier 1 storage, IBM is still going to be "pushing tape forward" as the way to go, Andrews said.

That's right - IBM's answer to enterprise-class solid state flash drives, the game changing, highest performance, fastest random-access, permanent on-line storage innovation to hit the Information Storage market since the hard disk drive is...you guessed it - TAPE!

OK. Calm down. Stop laughing so hard. You'll hurt yourself.

Seriously. STOP it! Right now!

Don't make me pinch you...

TAPE, he said. TAPE!

the ds8000 is now only for mainframes

But perhaps the most telling thing of all is something that both Charlie and IBM's brand marketing blogger TonyP independently have confirmed over the past couple of days: the near-death DS8000 is basically only being kept on life support in order to meet the needs of mainframe customers.

But Andrews said there are reasons to keep the DS8000, such as its mirroring features and a performance profile suited to transactional processing, rather than the archival and file-sharing applications IBM is planning for XIV.

See, unspoken is the fact that mainframes don't really need storage partitions - the zSeries in fact provides all the partitioning you need to insulate applications from each other. And with the z10 supporting over 336 FICON ports, you're going to need multiple storage arrays anyway to support the I/O load, so partitioning is pretty much moot. Same for thin provisioning - mainframes routinely operate with storage utilization that would embarrass any open systems storage administrator.

So it's pretty clear that since the only enhancements still being made to the aging DS8000 are in support of IBM's mainframe monopoly, and the open systems storage market is left to be served either by re-badged NetApp filers (aka N Series), or by IBM's OEM'ed mid-tier storage products, or perhaps (eventually) an as-yet undefined XIV-derivative.

A fact that is blatantly underscored by sir TonyP's latest blog post in IBM's strategy for the future data center:

...the IBM DS8000 can focus on its core strength, managing databases and online transactions for the mainframe.

Given all the compute power wrapped up in the latest z10, I guess the target customers for IBM's latest mainframe 'as a group' must have unlimited power and relatively light I/O workloads, since their only storage choices are a DS8000 with Fibre Channel hard disk drives whose performance remains locked in the 90's.

No blazin' flash drives with the lowest watts/IOP on the planet. And no cost-effective SATA drives for the lowest $/GB and watts/GB for on-line storage.

Oh, that's right. They can use TAPE.

the best storage for ibm mainframes comes from emc

Thankfully, there is an alternative to the abhorrent lack of attention that IBM is giving the enterprise storage customer.

As EMC demonstrated all this week at SHARE, IBM customers can leverage state-of-the-art storage technologies to maximize performance, efficiency and economics - but only if they're using Symmetrix DMX. And they can do so with assured compatibility and interoperability with all the new z10 features, thanks to technology licensing and cooperative support agreements between EMC and IBM.

As a result, today's Symmetrix DMX-4 stands alone, ready to meet the needs of both IBM mainframe and open systems hosts (simultaneously, if so desired).

With the DMX4, you can mix 10K & 15K rpm Fibre Channel drives alongside 1TB SATA drives to optimize price-per-gigabyte for less performance-intensive information.

DS8000? No hablo SATA

And whether for mainframe online transaction processing, or for those cache-hostile open systems applications that IBM suggests you host on the z10, flash drives in a DMX-4 can deliver read miss response times as much as 10x better than the fastest hard disk drive.

DS8000? No hablo SSD (use Tape!)

And as I noted yesterday (in revenge of the mainframe), if you want the fastest "tape-less" nearline processing possible for your z10 (or earlier), the EMC Disk Library for Mainframe sets a new standard for off-line/nearline storage.

IBM? There's no such thing as a tape-less VTL.

So for all you IBM mainframe customers watching the final gasps of breath from the DS8000, and wondering where you're going to get state-of-the-art enterprise-class storage to go with your shiny new z10?

Not to worry, EMC has been listening to you (as a group), and they have just what you need. Now.

it's all about the white space

But ole' Charlie couldn't leave well enough alone, it seems, for he wraps up his comments with an admonition that IBM's recent XIV acquisition isn't about replacing the DS8000 at all (as I had previously suggested).

No, according to Charlie,

"...XIV was built to fill in a 'white space' in the market where we didn't have an offering, not to replace something we already have."

Seems to me that I may have been right after all - Charlie's statement is truer every day, as the DS8000 addresses less and less of the requirements of enterprise storage customers, leaving plenty of room for XIV to (try to) fill...

You just can't make this stuff up!

 

[UPDATED 7AM EST February 29th, 2008]

Last night TonyP added a post detailing the DS8000 enhancements for the new System z10 EC.

Turns out that all the noise about there actually being enhancements made to the DS8000 was perhaps a bit premature. At least 2 or 3 of the 5 enhancements are actually features that will off-load work off the DS8000 back into the z10. I guess the zSeries engineers got tired of waiting around on their storage platform to get things done.

And in a rare bit of truth-in-blogketing from the IBM spin machine, TonyP actually admits that 4 of the 5 enhancements he describes actually don't even have a price, packaging or a delivery date!

You gotta admit, it takes Weekend At Bernie's -like chutzpah to go to these extremes in an attempt to make it look like the DS8000 is still alive.

 

technorati tags: IBM,EMC,DS8000,Symmetrix,DMX,DMX-4,enterprise storage,flash drives,solid state storage,SSD,mainframe,z10 EC,z/OS,VTL,virtual tape library,disk library,SATA,tiered storage,enterprise class,vision,strategy,EDLm,nearline,XIV,Nextra

Comments

[Chuck Hollis]

The "Weekend at Bernie's" reference is PRICELESS!!

Now I'm ROTFLMAO...wait, whew, I'm OK now.

And, when you think about it, it's turning into a comedy. I just worry about their customers who don't get the joke here.

[Scott Waterhouse]

Barry... I am not sure I see this as any more than the traditional split brain scenario that is afflicting Sun and IBM with respect to the role of disk vs. tape. In a recent debate with TonyP, he basically admitted that all 4500 VTS systems they have sold are mainframe attach. So perhaps they have made nothing more than a business decision that it is better to prop up their VTS business than it is to try to move functionality to a nearly comatose DS8000 platform? I mean you and I know that doesn't address the requirements of enterprise customers for ultra-high performance, but... just like they are adding storage functionality to the core Z/OS platform, maybe they have decided to leave tiering to DS8000 + VTS? One is crippled and almost irrelevant, one more or less healthy. Which one would you invest in?

[Barry Whyte]

So what is it that makes it dead? Have we stopped selling it? No. Have we stopped adding features? No. Have we prioritised those features. Yes. Will there be more features later this year? Yes. Did we just sell more in 4Q08 than any previous quarter? yes. Have we had any customers pull them out and replace them with DMX just so they could put some really slow SATA drives in their shop? No.
Does it perform badly? No. It had a respectable SPC-1 benchmark - at least we admit what it will do - and don't have to hide behind SLA's that mean we'd ship in another entire frame if it didn't meet requirements. You can judge it up front and get some real data.

So all in all, I'm struggling to see exactly how its dead or dying? Have you got anything else other than SATA (no real need for monolithic tiering when you have a virtualization solution that works) RAID-6 (which is only really needed when you have SATA) 4Gbit drives - which really and truly are a bit of a wet fish - since no drive can actually sustain more than 2Gbit in MB/s (200MB/s) and any FC/AL loop topology / connections to the controller are likely the limiting factor way before the actual drive interface itself.

For more on my thoughts of the interview you quoted from (I've been asking some questions internally btw) see my recent post.

[the storage anarchist]

BarryW -

Hey - I'm just cuttin' and pastin' here...I was very careful not to quote Charlie out of context or anything.

You ask what's missing from the DS8000? How about the things that Beth called out (and I repeated up above): variable sized partitions (promised since 2003), thin provisioning (still not publicly committed), and flash drives (use TAPE?).

Or how about something more than a measly 8GB of pending write cache? Or more than two partitions, ones that can be dynamically reconfigured without restarting the box? Or a drive larger than the measly 500GB FATA (LC-FC) the DS8000 supports today - you started down the big-and-cheap path...what happened, didja get stuck without a drive roadmap or something?

As to "no drive can sustain more than 200MB/s" - I agree, no DISK drive can. But how about Flash Drives? Do you think they'll handle more than 200MB/s any time soon? Maybe?

Or are you only going to stick Flash drives behind an SVC? Or perhaps wait for NetApp to integrate them with the N Series for you?

Or maybe you're just going to put them in the XIV Nextra, as Andrew Monshaw seems to be telling folks this week?

Either way, it looks like Flash Drives aren't destined for the DS8000 any time soon ("tape," he said. TAPE!)

BTW, seems kinda funny that Tony, Charlie and Andy can't seem to agree on what to say publicly...

[Barry Whyte]

Andy's interview could be misread to mean only XIV, but I think that a case of 'article editing' rather than anything else. For the record, IBM's flash investigation is most definitely not limited to XIV. Where and when, or what will become clear in due course.

Sure, Flash is bound to go beyond 200MB/s, just keep adding more channels and you are sorted. But FCAL is probably not the way to go with that - far to much overhead with the arbitration.

[Ron Lloyd]

Barry, I think that you missed the point. They plan to use DUCT tape to hold the product together through the next product cycle. You thought that he meant magnetic tape, which obviously makes no sense at all.

[Nigel]

You want to be careful Barry.... TAPE might turn out to be an acronym (Totally Amazing Performce Enhanced drive.........). Or even an internal code name for a top secret SSD killer project within IBM.

They might be feeding you all this fud and watching you take it while gearing up to rock the world with with the DS8000 series superduper turbo injected intercooler model.

Or may be not.