« 1.041: Seagate drops SSD patent lawsuit against STEC | Main | 1.043: the fine art of strategy »

February 25, 2009

1.042: modular storage - what's in a name?

modular storage OK. I did the last one. Now it's your turn.

What is the definition of modular storage?

It sure seems that I really got under the skins of the Dancing Giraffes over at NetApp with my "Flash Dance" expose of how slowly the competitors are embracing the value of flash technology.

First there was their uninformed challenge to the term Enterprise Flash Drives. Not surprisingly, everyone else (except NetApp) seems now to understand that indeed there is a sufficient differentiation among solid-state storage devices to justify the "enterprise" classification, just as we also distinguish enterprise disk drives.

The second salvo from NetApp comes in the form of a brandy-new blog, authored by a pool of NetApp engineers, although it seems Mike Riley has taken the lead role, authoring the first 3 posts. In the latest post, Mike seizes the opportunity to a) cast me as an angry villain, b) offer me a hug, and c) assert that the last economic downturn fostered the era of "modular storage" and the end of monolithic storage's world domination.

Hence my question: what the heck is "modular storage" (as you can see from that link, even Wikipedia doesn't seem to know).

Mike continues with an outlandish assertion that "Hardware offers zero differentiation".

I guess I can agree with Mike in the sense that we all can use the same parts to make our products. But there does seem to be rather significant differentiation based on what components we actually do choose to use.

For example, the fact that Symmetrix DMX4 natively supports EFDs today, while NetApp still doesn't support flash drives of any sort with their mainstream storage arrays is pretty highly differentiated.

EMC is delivering the value, while NetApp is bringing forth more people rappers dancers.

massive modular storage

And seriously, I am interested in hearing how you'd characterize or define "modular storage" these days…does the term even have meaning any more?



Oh, and thanks for the hug, Mike. You'll never know how much I needed it this week!



TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d834c659f269e20112790ce4bf28a4

Listed below are links to weblogs that reference 1.042: modular storage - what's in a name?:

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

marc farley

I love the photo at the top. LOL.

Martin Glassborow

Drives...can that be a Modular SSD array please? I've got a few from vendors! And if I install the NetApp Simulator on them, can I do press release that NetApp now support SSDs? Or perhaps I could label the USB hub as a storage virtualisation appliance?

Jimmy Floyd

Hi,

I will be involved in purchasing a new array soon. And it won't be from EMC or Netapp. You're both missing the point these days!

But blog away chaps, this posturing is great fun!


P.S. Even though Netapp will be getting a sale, SanScreen in, ECC out, Solutions Enabler for provisioning.

Regards,

Jimmy.

the storage anarchist

Glad you're enjoying the fun, Jimmy. Life's too short...

But not sure which point you think we're missing...please elaborate!

Owen

The general usage for the term "modular array" is one that is build from modules installed in a standard 19" rack. Exampled include

- EMC Clariion
- NetApp FAS
- IBM FC4xxx/Sun/LSI
- HP EVA

When the storage is shipped by the frame, it isn't modular

- EMC DMX/Symm
- HDS USP

I agree that it is a pretty silly term, I would class the USM-VM as modular (12RU is a big module), but it runs the same code as a USP.

I actualy like (but don't fully agree with) the comment ""Hardware offers zero differentiation"".

Bottom line is there isn't a lot of difference in the hardware used by the major players. At the hardware layer, what is the real difference between and EMC Clariion and a NetApp FAS? Slightly different I grant you, but basically the same kit.

The big difference is in the software (Firmware, management layer, whatever you want to call it).

the storage anarchist

So, why is a 19" rack form-factor the distinguishing feature? I'd venture that the overwhelming majority of CLARiiONs are shipped pre-installed in an EMC-supplied rack. And a DMX4-950 also uses the exact same cabinet/rack as does the CLARiiON line - doesn't that make it modular too?

And then, the entire DMX4 line-up is constructed of the same components that make up a DMX4-950, mounted in a slightly larger rack that accomodates additional power supply and battery backup modules. Would you consider that the same parts + different cabinet = modular as well?

I'm getting the feeling that any definition of "modular storage" is inexact, and used today only out of convenience rather than as a true classification.

Owen

"""I agree that it is a pretty silly term"""

The term is bluring, no doubt about it.

I first came accross the term when it was used by people selling frame based arrays, I think at the time the terminology was "enterprise" vs. "modular". As all the Frame based array vendors also had modular storage, they needed something to clarify the two architectures (both EMC and HDS have different archtectures for their Modular and Frame based arrays). I havn't played with one of the baby DMXs, but HDS have branded they Baby USP with a M tag for modular.

NetApp don't sell a "Box" that compares to a USP, DMX or DS8000. They sell controllers and sheles full of disk, like a Clariion, AMS or EVA. You expand these by buying a 3RU shelf of disks, not by adding a frame.


Alex

Hmmm....and here I thought Modular vs Enterprise were differentiated by 3 things:
1- Mainframe support.
2- Cost ($$$).
3- Features.

IMO, the biggest differentiator on Modular vs Enterprise is cost. Feature-wise, some 'enterprise-level' features (thin provisioning, SSD) seems to be pushing down to the midrange level due to customer demand.

Mike Riley

Yep - I'm the brand new NetApp blogger on the block! I'd refer you back to my post because I think we're getting sidetracked on the definitions of modular and monolithic - not the point of the post. (BTW, Barry, rev up your Wikipedia search again and it will take you about 15 additional seconds to find a good categorization of modular and monolithic arrays. Better yet, run the search "modular storage" on EMC's website and it will turn up the CX4 product page. Somebody at EMC seems to have an inkling as to what "modular" means. I'd ask around.)

The point of my blog was to use these categories to show that hardware technology doesn't drive adoption (especially, drive types). Economics do. If you look back to 2000 - 2001 there was a concerted effort by EMC and others to protect their frame array sales. The argument made was based on an alleged superior hardware design to that of modular storage (as we generally define it). Well, based on customer sales, we know that the hardware superiority argument didn't hold up. Modular storage has greatly infringed on what was formerly considered the domain of frame arrays. So, the premise of your SSD blog is faulty and history bears that out.

Now, when it comes to qualifying new drive types? SSD is just another drive type. On one end of the spectrum we all qualified SATA; now on the other end of the spectrum we've all qualified SSD with SAS and FC in between. Now what? I think it all boils down to what we do with them and that has to do with the software. That's where you'll find the differences. That's why we've found success front-ending other people's storage with our controllers. That's why we have space saving guarantees using other people's disk. It isn't the disk and people know it, so they're more than willing to entertain how to get the most out of that commodity on their floor.

You may disagree with how each vendor has integrated a new technology but the fact that every vendor has it pretty much makes it a commodity. So, if you want to stand on the hardware is a differentiator argument, I say more power to you. We'll continue down the path of software as an enabler using any hardware - yours, ours, IBM's, HP's - and we'll have to see how it shakes out.

the storage anarchist

Alex - that definition doesn't make any sense to me - if it costs more, it's not modular? And if it's inexpensive, it IS?

C'mon, we can do better than that, can't we?

the storage anarchist

Mike -

So when a 146GB EFD gives up to 30x the performance of a 146GB 15K rpm disk drive but costs only 8x more, wouldn't you say that the economics are indeed driving customer decisions?

And when you no longer have to stripe data across dozens of spindles to meet your IOPs and response time objectives, don't you think customers will actually SAVE MONEY with EFDs?

And unless I've missed a press release here, I still haven't seen NetApp qualify any EFDs in your Filers. Your customers can put SATA drives into the FAS's, yes, but if they want to reap the ECONOMIC BENEFITS of EFDs, no dice. (And no, PAMs don't count, because you're still only shipping DRAM-based PAMs).

So you're right, and I'll even agree with you - economics will indeed what drives change in the market. Right now, the divide is between those that OFFER the economic benefits of EFDs, and those that Don't.

Clearly, NetApp is in the "monolithic" camp this time.

And please, don't be bringing your space-saving marketing gimmicks around my blog. Every sane storage admin knows that they'd have to disruptively migrate their data into a totally new operational and management paradigm in order to play your game. Utilizing the snapshot and thin provisioning capabilities of their existing platforms will cost them far less in capital, operational and environmental expenses. Nice marketing ploy, but without any real ROI we both know that the smart customers won't bite.

 Mark

Quite frankly I really don't think NetApp has the right to talk about efficiencny. By the time I stopped piling on the overhead using my calculator my fingers were sore. Please do not attempt to educate me on the NetApp way of calculating, I already know that one. Make the solution match the competition, come in later with several shelves under your arm and say oops. Also the definition of Modular should be, NetApp. NetApp is not Enterprise. NetApp has hit the wall on many fronts and is anxiously awaiting for the "convergance" to come to fruition. Without its arrival and subsequent success, its Modular defined.

Dwayne Sinclair

Mike – In my 20 years of enterprise experience, cost was only one metric in a decision analysis matrix. Likewise, frame vs modular is a packaging discussion and not a “hardware superiority” discussion. Lets get back to the two issues we all face – 1. What are the business problems I am trying to solve and 2. Building scalable/adaptable enterprise IT solutions.

Dwayne

Ps. I’m not a big fan of leading tchnical discussions with suppositions.

Mike Riley

Team, I'm glad we at least agree that economics drive technology adoption and Dwayne, you and I are in violent agreement. If you check out my blog, I say exactly that. It can be Campbell Soup cans and kite string on the back end - as long as it meets the customer's business requirements - great. I'm not sure what you mean by "suppositions" though. I'm merely looking back at history to find out what shifts occurred if any in the storage market during the last recession and there was a significant shift from frame arrays - USP, DMX, DS8000 - to modular arrays - Clariion, NetApp, HP, 3Par (examples). You have to ask why? Prior to the last recession, customers had expanding budgets and what they were using worked and worked well. If lowering costs (acquisition and ownership) was your proposition, it didn't do well. That proposition was basically attempting to solve one of the only problems customers didn't have. It wasn't until IT budgets began to contract that customers started to search for alternatives. They were looking for ways to streamline costs without compromising on enterprise performance and data protection. What they found was that the modular storage offerings - and I mean as a storage market, not just NetApp or EMC - could provide the level of service they needed for a significant portion of their work. Customers exploded the myth that hardware packaging somehow made one array Tier 1 and another Tier 2. Gone were the days of "if you want the Cadillac, you have to pay for the Cadillac" (an old selling phrase used on me when I was a customer). Despite the concerted efforts on the part of frame array sales people, the hardware commoditized significantly during the last recession.

I suggest it's the firmware/software lashing this commodity hardware together that is the true differentiator. Software is responsible for delivering all of the scalable/adaptable (and I'll add resilient) features customers demand. There are plenty of SSD solutions out there that enterprise companies will not use primarily because they do not support the resiliency features needed for a production environment.

When it comes to this dust-up around SSDs, merely qualifying a new drive type in and of itself means little to the customer. Barry, in your own example, you're simply substituting SSDs for FC drives. There's no innovation in that at all. Now that many of the major storage vendors have qualified SSD drives (the NetApp press release you may have missed is on Texas Memory Systems), you have a defacto commodity on your hands since based purely on qualifying the hardware, we can all make the same claims of the economic benefits. I can find your opening sentence in every vendor's ppt on the using SSDs. Just swap out the vendor name in the title which to a customer makes this a commodity - 8 vendor talking heads all chanting the same mantra. Now the question becomes what sets one apart from another? I suggest it's how you leverage the hardware and that can only be done via software.

And, I do respectfully disagree with you Barry on the idea that you can pick and choose your solid state devices based on whether they are DRAM or flash. Again, you're making a hardware-based argument. The goal isn't to employ SSDs or flash or DRAM. The goal is to figure out how best to leverage commodity hardware (of which SSDs, flash, FC, SAS, SATA, DRAM are examples) to solve the customer issues Dwayne points out. Software does this and increasingly, it doesn't matter who the backend storage device is and I think that's a bit alarming to those who do think innovation is based on hardware integration.

Mike Riley

Oh - and Barry - we have multiple Fortune 500 customers that actually do use us in front of your storage (National Semiconductor is one) where we can and do flip on features like dedupe and thin provisioning. It's not marketing if you actually do it.

And, Mark, if you are willing to do an apples-to-apples comparison on usable space, I think we stack up very well. I won't attempt to educate you since you say you don't want or need it. I'll just say that I respectfully disagree with your assumptions. The beauty of math is everyone must follow the same rules so the numbers are what they are.

As far as not being enterprise, well I again would respectfully disagree. I don't know what your definition of enterpise is, though, and NetApp doesn't fit into every circumstance (e.g. mainframe). I will say that I've talked with customers and asked what happens if our systems go offline and they say "people die" (literally) or trades do not go through or planes are grounded, then I have a pretty good understanding of what is required of us. I don't know if we meet your specific definition or not. We may not.

the storage anarchist

Thanks for the response, Mike.

Just a couple of nits: first, if it were as simple as qualifying a new device to get SSDs into an array, I hardly think it would have taken everyone else more than a year to do so...and in fact, NetApp is among those who apparently still have not figured out how to get SSDs qualified. I have no idea why you would intentionally delay qualifying the STEC EFDs in your Filer products while qualifying the external SSD array from TMS - it seems that if you think customers want to use EFDs as a tier, you'd want to offer support for both your vFiler customers as well as your (significantly larger) integrated Filer base.

Truth be told, integrating EFDs requires far more than a simple qualification. IBM's white paper lists a lot of the "excuses" (see my post of March 7 for some insights). Getting to the benefits of performance and response times of EFDs probably challenges most architectures, and I could liist of several key modifications to Enginuity that have been intentionally made over the past 4 years so as to enable the efficient use of Flash - I could, but I won't, because it's more fun to watch all those other 8 vendors figure it out for themselves.

As to the definition of "modular" - I agree, basically, that it really has nothing to do with "modularity" any more. High end arrays and mid-tier ones today are all highly modular. And indeed today, the real price differentiation for "enterprise" vs. "mid-tier" is far more related to the software functionality differences than to the old monikers of "monolith" vs. "modular," much less any differentiation in the hardware (although technologies that allow "high end" to start small and scale up performance, ports, memory, AND capacities are still a differentiating factor).

Robbie

NetApp Storage Efficiency Guarantee debunked
http://snurl.com/dh67j

John F.

Barry,

You're forgetting about economics. Comsomer decisions are besed on the solution fitting the need, and the then the bottom line is cost. Why would you introduce a technology when the cost is 3 orders of magnitude higher than the alternative? When cost is declining, approaching 2 orders of magnitude higher and dropping, considering lead times, etc. well that would seem to make a lot more sense (unless you don't care about cost). Being first isn;t as important as being cost effective.

the storage anarchist

John -

You and your fellow NTAP pals can keep trying to recraft the facts, but EFDs have NEVER been "3 orders of magnitude more expensive" than disk drives.

Today you can get EFDs from EMC for less than 10x the $/GB of a 15K rpm drive, and get 30x (or more) IOPS as a bonus.

Stick with me here: Cost is indeed the key factor, and when you can support your application's IOPS requirements with 4 EFDs instead of 120 15K disk drives, the EFD solution IS CHEAPER!

Chuck's coverage of EMC's Strategic Forum provides more details on how EFDs can actually save money.

Mike Rley

Since I droned on the last two posts let me say that I agree with Barry on more than a few things here. You can make a storage efficiency/cost case for SSDs. You do have to factor in the performance angle but, then again, that makes sense since SSDs are targeted at that high IOP/low latency workload. On a per drive basis - yeah, the cost is 3X - 8X greater but you do only need a fraction of them for these workloads compared to 15K FC (or SAS) drives. We're not going to target these devices for archive storage although with some intelligent caching to SSDs, you can make a great case for cheap and deep at the core with cache acceleration elsewhere (e.g. in the fabric, in the controller, as part of the disk architecture).

On the one nit (and it really is): STEC or TMS question. With ONTAP 7.3.1 we can mix and match storage from various vendors behind a NetApp controller. We essentially V-Series the whole lot now. The lines of what is "integrated" NetApp storage and what is not are blurring. We'll continue to ship NetApp shelves but as new opportunities come up you can see the value of front-ending other reliable disk arrays. Ultimately, that's plumbing. The higher order value is figuring out how to help the customer with their business issue. Honestly, they could really care less as to how much sweat equity we (EMC, NetApp, IBM, HP) have to put into all of this to make sure the plumbing doesn't leak. Just don't leak is the right answer.

The comments to this entry are closed.

anarchy cannot be moderated

about
the storage anarchist


View Barry Burke's profile on LinkedIn Digg Facebook FriendFeed LinkedIn Ning Other... Other... Other... Pandora Technorati Twitter Typepad YouTube

disclaimer

I am unabashedly an employee of EMC, but the opinions expressed here are entirely my own. I am a blogger who works at EMC, not an EMC blogger. This is my blog, and not EMC's. Content published here is not read or approved in advance by EMC and does not necessarily reflect the views and opinions of EMC.

search & follow

search blogs by many emc employees:

search this blog only:

 posts feed
      Subscribe by Email
 
 comments feed
 

 visit the anarchist @home
 
follow me on twitter follow me on twitter

TwitterCounter for @storageanarchy

recommended reads

privacy policy

This blog uses Google Ads to serve relevant ads with posts & comments. Google may use DoubleClick cookies to collect information (not including your name, address, email address, or telephone number) about your visits to this and other websites in order to provide ads about goods and services of interest to you. If you would like more information about this practice and your options for not having this information used by Google, please visit the Google Privacy Center.

All comments and trackbacks are moderated. Courteous comments always welcomed.

Email addresses are requested for validation of comment submitters only, and will not be shared or sold.

Use OpenDNS