Hitachi's New VSP: Separating The Wheat From The Chaff

Well, we all knew that Hitachi (through HDS and HP) would be announcing some sort of refresh to their high-end storage platform sooner or later.

As EMC is Hitachi's only viable competitor in this part of the market, I think people are expecting me to say something.

Well, rather than get into the weeds, I saw a few aspects of the announcement that made me realize that both companies see the world very differently.

To Begin With

If you're a high-end storage kind of person, your universe is basically a binary star: EMC and Hitachi orbiting each other, with the interesting occasional sideshow from other vendors trying to claim relevance in this space.

Demand for these high-end storage architectures appears to be on the upswing, as far as I can tell from customer conversations.  Smaller and more modular storage architectures have their roles (and fans), but more people are coming around to the unarguable advantages of these purpose-built scaled platforms.

Given that it's basically down to two vendors in this space, it's getting progressively easier to spot the differences on how each sees the world.

Disclaimer

I have only made a cursory investigation of the HDS announcement materials at this time.  I work for a competitor, so everything I say should be taken with the appropriate caveats.  And I have worked in and around EMC's Symmetrix business since I got here 16 years ago, which has resulted in an unavoidable bias.

The Question Of Processors

EMC makes no bones about it: we see the future as Intel-based -- not only for the vast majority of server workloads, but for building advanced storage products as well.  As a result, the Symmetrix VMAX (as is true for all other EMC storage products) is Intel-based, rather than the more proprietary hardware that is powering the Hitachi offerings.

On a tactical level, EMC is betting that the well-known tick-tock of the Intel roadmap will give us more processing power and more cores (at lower power and associated costs) than any other alternative.  On a strategic level, we also are betting that Intel-based storage platforms will be able to gain new functionality more quickly, simply because we can leverage the vast Intel-compatible ecosystem.

It's a bold bet.  We're basically claiming that more customer value can be derived by investing in software, and riding the wave of ever-improving industry standard hardware components to build our hardware platforms.

It'd be relatively easy to claim that "storage is different", and make a case for continuing investment in specialized ASICs and processors and the like.  Well, in my opinion, storage arrays really aren't all that different -- they're basically huge server engines that do I/O for a living.  

And, just like the market attractiveness of a server built with proprietary parts is somewhat limited, its inevitable that the same will be true for storage arrays.

[Whoops!  Realtime update -- Hu Yoshida just mentioned during their announcement that -- yes -- Hitachi and HDS are now part of the Intel fan club.  Welcome!]

Single Footprint Vs. Federated

Much can be made about who's got the biggest array: capacity, spindles, ports, memory, etc.  And there's something inherently attractive in "bigger equals better" positioning.

But in the real world, there are pragmatic limits to how much storage you want to put in a single physical enclosure, or -- more realistically -- a single physical location.

The current design theme from Hitachi appears to be pretty clear: a single, monolithic array with as much stuff as you can cram into it, or -- in some cases -- put behind it.  

By comparison (although the VMAX specs as a stand-alone array are nothing to apologize for) the underlying design theme is very different: federated pools of storage resources, working together to create a fabric of capabilities (think VPLEX and progressively more distance), rather than putting it all in one ginormous array.

Storage As Part Of The Enterprise Stack Vs. A Stand-Alone Domain

EMC sees storage as a set of intelligent resources and services that expose themselves to other layers in the enterprise stack.  A good example of this thinking is the incredible levels of investment in VMware and other hypervisors, not to mention significant investment in the new generation of converged IT stacks (think EMC Ionix UIM and Vblocks).

In this world, the role of storage management and administration is fundamentally different: it's about delivering services carved from pools of resources that others consume, rather than a self-contained world of storage along traditional boundaries.

Call it cloud, call it IT-as-a-service, call it whatever -- we're pretty adamant that the future of storage is integration, not isolation.

The Differences In Thinking Are Clear

I'm sure there will be much back-and-forth between the two vendors regarding the specifics of our respective storage arrays.  There's a lot at stake, and most of the momentum (as well as customer investment) has been going to EMC over the last few years.

Hitachi will need to make a case that -- not only is their product competitive -- but that they've made the right bets on the future of high-end enterprise storage, and -- more importantly -- the context it must work within.

And, by that measure, I think they've still got some work ahead of them :-)

Comments

[Chris M Evans]

Chuck,

One quick comment; although the core of the VMAX is Intel based, the most dependent piece of VMAX is the Virtual Matrix Architecture, otherwise you can't scale, So is that Intel commodity hardware?

looking forward to some good discussions.

regards
Chris

[Chuck Hollis]

Hi Chris

I don't understand your point. Are you saying that -- because there might be an ASIC or two in the current array -- that the entire premise isn't valid?

-- Chuck

[Chris M Evans]

Chuck

I'm saying that everyone's architecture uses custom components. VMAX may use intel, but so does VSP. Both platforms also use custom components, so the intel discussion isn't relevant.

Chris

[Chuck Hollis]

Whoops -- updated my post to reflect that -- yes -- Hitachi announced they'd be using Intel processors going forward :-)

-- Chuck

[Visiotech]

For more about HDS VSP look here
http://blog.nigelpoulton.com/

[nate]

I don't know.. "purpose built" and "running on Intel cpus" just doesn't sound right..

When thinking about those ultra high end storage arrays in the form of the XP/Symmetrix etc for some reason I can't help but think of some comments AMD made last year about the 8-way server market. Where they said they were getting out of that market because it has collapsed over recent years with the rise of multi core CPUs, and accounted for less than 0.2% of processors shipped per quarter(vs 4-way servers at 4%).

Funny that just after AMD exits that market, HP launches their first 8-way Intel server in(if I remember right, it was a Compaq system) 7-10 years, all 8-way servers from HP at least until recently were AMD based.

I'm sure there'll probably always be a market for such high end systems, probably like there'll always be a market for IBM mainframes(what a coincidence that those systems have mainframe connectivity)..well you may see where this is
going :)

With the types of companies I work for (small, sub 300 people), unlikely I'll ever get exposed to such a system directly.

I ran into a couple HDS people at a bar a few weeks ago they were pretty excited about this new VSP and suggested I check it out, well while it certainly looks fancy, it's not really in the running for my pending consolidation project.

I was told a funny story about this guy I know in the area about HDS. Some HDS folks came out on site to his company and asked them why he liked XX storage array more than HDS, and he said among other things it's easy to use. And their response was - well if you attend our 3 week training course you'll probably feel just as comfortable on our system as you do on theirs. He said - you're kidding right, you just made my point.

[Nigel Poulton]

Hi Chuck,

Interesting take.

One of the challenges with Hitachi (I see the merits but also the challenges) is that they generally dont announce things before they ship them. So they may very well be working federation behind the scenes, and dare I say even ship VPLEX like functionality within the VSP before EMC get in to to VMAX (assuming that is the EMC strategy). Crazy thought but you know where Im coming from.

When I raised the very point yesterday on Twitter, about federation being better than single large array, Roberto Basillo of HDS said that they are working on that.

Personally I'm gald for the competition in the e terprise space, keeps folks innovating and keeps thins interesting.

Nigel

[Az990tony]

Chuck,
Given that IBM is #1 for high-end disk storage with FICON-attachment, perhaps you meant that EMC and HDS are a binary star of peers competing for second place?

Tony Pearson (IBM)

[Chuck Hollis]

Hi Tony

No, I was quite correct in my assertion.

Many of us in the storage industry believe that IBM currently does not field a competitive high-end storage platform. IDC market share numbers bear out this assertion, as you probably know.

For those of you who might be unfamiliar with the terminology, FICON is an IBM-specific mainframe storage protocol that bears strong resemblance to FC from a transport perspective, and ESCON from an architectural perspective.

Unless you happen to own an IBM mainframe, you've never heard of it, nor would you ever really care.

It's reasonable to expect IBM to sell a respectable amount of storage with their mainframes using a protocol of their own design -- although IBM's two competitors in this rather proprietary space (notably EMC and Hitachi) sell more together than does IBM.

Tony is also making the (presumably intentional?) logical error of confusing storage platform architecture with the attachment protocol.

I'm assuming he knows the difference. I certainly hope so, anyway.

-- Chuck

[Chuck Hollis]

Hi Nigel

Nice to know that Hitachi acknowledges the point that there are pragmatic limits to how much stuff you can reasonably cram into a single physical enclosure.

And no doubt they're working on something along the lines of VPLEX-style federation, although I'm reasonably confident that they may find it a rough go, especially when we get to multi-node topologies separated by considerable distance.

One thing we both agree on: competition is good for all of us. Frankly, a lot of us wondered what had become of our friends at Hitachi (and HDS).

Nice to see they've still got some game left in them!

Those guys at IBM, however --

-- Chuck

[Josh Krischer]

Chuck, with all the respect I have some remarks/objections to your analysis:

1. High-end storage arrays differ a lot. Those who try to commoditize them either lack sufficient technical skills or mask the inferiority of their own products. The input and output are the same but functionality, availability and performance differ a lot.
2. You are neglecting the DS8x00 from IBM. Since October 2007 it has full functionality and the pace of introducing new functionality is the fastest in the industry. The DS8700 has Easy Tiering (advanced sub-LUN automated data placement) since April, the first in the high-end storage. EMC announced FAST II on VMAX in April 2009 but still on paper only. EMC neglected the competition in 2001and we both know the results.
3. A tailor-in technology will we always more effective than a subsystem built with standard components. The difference is in cost of manufacturing but if it is sold at market price .The benefits are in vendor margins but these are not passed to the end-user. For end users what is important is the get points above at acceptable price. Your discussion on “question of processors” is a simple “red herring” deviating the discussion from essential issues to irrelevant comparison. In addition to the Intel processors, VSP uses vast amount of ASICs and other processors specially designed for the VSP.
4. Single Footprint Vs. Federated- if you read EMC marketing material you will find that since many years EMC claimed the highest scalability, despite the fact that the number of host connections was in most cases two small for full populated subsystem . I know several full populated USPs from Hitachi but I didn’t hear about full populated DMX or VMAX. Have you read “The Aesop fable of The Fox and the Grapes” or the comparable La Fontain story?
5. Storage As Part Of The Enterprise Stack Vs. A Stand-Alone Domain. HDS has been cooperating with VMware since 2002, becoming a VMware Premiere Global Partner in late 2006. Their alliance as resulted in comprehensive interoperability via certifications and qualifications. The cooperation was initially focused on integration of HDS storage subsystems with VMware Site Recovery Manager and later expanded to create tight integration in initiatives such as vStorage and vCloud. This integration leverages storage subsystem functions such as thin provisioning and dynamic resource management in VMware environments. Anyhow, in my opinion EMC should be very careful not to misuse the benefits of VMware ownership in an open world; you may kill the goose that laid golden eggs.
Hitachi’s Storage Command Suite combines advanced data management and protection capabilities in a simple, common (GUI and CLI) user-friendly centralized interface to manage the Hitachi storage infrastructure, whether it is comprised of virtualized heterogeneous storage or non-virtualized homogeneous storage. All other large storage vendors use different storage management applications for each platform as opposed to this single interface which manages the complete Hitachi Data Systems storage portfolio, from the high-end VSP to the HCAP archiving platform. Users are able to see all their heterogeneous storage assets as a unified pool as opposed to managing disparate islands of storage.
Can EMC manage the VMAX, CLARiiON, Centera and Celera with a single management program? A federated eDiscovery?
6. The Differences In Thinking Are Clear I agree: Hitachi choose the way to continue to develop state-of-the-art storage technology tailored to current customers’ requirements and EMC decided to continue with standard industry components, mask that with powerful marketing and make bigger margins.
I don’t work for Hitachi, but I am an analyst with electronic engineering background (from the same technical university that many graduates which built the Symmetrix) and more than 40 years of experience in storage.

[Chuck Hollis]

Hi Josh

Before we really get into this, I would encourage you to be more transparent about your business model and how that might influence your statements here.

I took a quick look at your "published work" on your website here (http://www.joshkrischer.com/workP.html) and I have to say it is indicative of a particular business model.

I'll make a deal with you -- you share with us who is paying your bills, and I'll tackle each and every one of these points to my best ability.

Sound good?

-- Chuck

[Josh Krischer]

Hi, Chuck
My business model has a nothing to do with my opinions on technology.
I am advising to end users, speaking on storage events such as SNW , Storage Expo or vendor’s events. I am writing whitepapers for vendors however the vendors don’t dictate me what to write and it also doesn’t mean that what I write is not correct.
I wrote similar papers when Gartner but less detailed when paid my salary. Many of my papers in Gartner were purchased by vendors for reprinting or posting on their web.
Josh

[Chuck Hollis]

Hi Josh

Thanks for the reply. I would argue your core premise that "your business model has nothing to do with your opinions on technology". Folks in your line of business frequently have to cater to what either customers or vendors want to be discussed. If vendors pay you for white papers (and you are aware of that), it's impossible not to be encouraged to write white papers that vendors will want to purchase, right?

My business model is that I work for EMC. My employment status greatly affects what I say. I make no apologies for it whatsoever. It's pretty easy to see where I'm coming from, especially when you see that big, honkin' EMC logo on my masthead :-)

I don't think anyone would really be interested in a point-for-point rebuttal of your lengthy comment. Doing so might be an interesting exercise, but who might really care?

The market has spoken, so to speak ...

-- Chuck

[Oracle_Storage]

Hi Chuck,

I trust you're well, it's been a while since we last spoke.

At Oracle we quite agree with your view of the world: it's about the whole stack, a commodity architecture and software differentiating features.

BTW, when is compression and dedupe coming into the VMAX?

--Ludovic.

[Chuck Hollis]

Hi "Sun / Storagetek / Oracle"

Today, customers get compression, dedupe and a bunch of related information management features by using a Celerra gateway in front of a block storage array such a VMAX. If you've been following us, you'll probably be aware that the Celerra VG8 is the current SPEC 'bad boy" from a benchmark perspective.

The more interesting question is when will VMAX support these features natively? Probably long before your ZFS-based NAS devices can be used for serious OLTP :-)

-- Chuck

[Helmsley]

Hi Chuck,

Interesting point that you did not mention anything about NetApp in this blog, especially when discussing about federated storage design and integration with virtualization players like VMware and Hyper-V. Or are we just focusing on the millions of IOPs and fanciful technology here only? :)

[Chuck Hollis]

Hi Helmsley

I don't know if you're aware of this, but NetApp does not play in this particular storage market segment.

In addition to not talking about NetApp, I also didn't mention many other vendors who also don't play in this category.

My guess is that you haven't really been exposed to the world of high-end enterprise storage.

-- Chuck