1.013: more on enterprise flash

Sometimes I just can't keep my opinions to myself.

Last Friday afternoon while I was perusing Chris Mellor's Blocks and Files, I was struck by the discontinuity of messaging that is emerging around the "best" application of Flash drives. From the "flash can't write fast enough" doom and gloom of the relatively unknown wanna-bee Infortrend to the "flash is too fast for external storage" justification for Sun's focus on server-based flash, it struck me that uninformed readers were likely being confused, if not outright mislead by the contradictory misinformation.

So I dashed a letter off to Chris outlining my concerns and rebutting the misrepresentations of the real benefits that the right flash drive is already delivering today - despite the unfounded claims of others in the storage and server industries. And upon his request, I granted him permission to publish this as an opinion piece on Blocks and Files, which he has done under the title of Missing the flash point (a fitting play on words, IMHO).

N.B. - Normally, I would have posted this to my own blog, but part of my concern is that the misleading information is being distributed by sites like Blocks and Files. My letter was in fact originally written in a context beseeching Chris to present a more balanced perspective on the realities of flash technology. When he offered to publish my opinion outright, I figured the best way to provide balance was to use the same billboard, so to speak.

Anyway, I hope that my opinion letter helps expand the conversation about flash technology. In my opinion, there's nothing to be gained by trying to polarize this into an either-or discussion. In fact, I honestly believe that there are appropriate applications for Flash-based storage in both servers and external storage arrays. And judging by the applications I've seen flash targeted for by customers, the early adopters are going to have a pretty big competitive advantage on their competitors.

The bottom line is this - the sooner that more vendors stop quibbling over who, how or where flash is "best" and actually start bringing real flash technology to market in their respective products, the sooner (and faster) we collectively will drive up demand and drive down the costs.

And that, my friends, is the whole point...the flash point, as it were: leveraging Flash to cost-effectively bypass the inherent limitations of mechanical disk drives...

IMHO

 

technorati tags: flash drives,flash SSDs,enterprise flash drives,EFDs,Infortrend,Sun,Pillar Data Systems,storage performance,storage consolidation,NAND flash,SSD

Comments

[Jeff Darcy]

Driving up demand may or may not drive down costs eventually, but it will drive up *prices* with greater certainty and immediacy. I guess that sounds really good to some people, but not to end users.

[the storage anarchist]

Thanks for the comment, Jeff. I assume you're referring to the market dynamics of supply and demand.

Frankly, I haven't heard anything that indicates that SLC NAND Flash is supply-side limited. In fact, I've read that the existing NAND foundries can create either MLC or SLC (although apparently not at the same time). And there are reports of several additional NAND foundries coming on line in the near future.

Thus, I really suspect that the current SLC premium pricing could be eroded quickly with more demand. But I take your point.

Neat blog site - thanks for the link. Sorry to hear about your tree...

[Pete Steege]

Absolutely right. Are disk drives only good for one type of application? SSD will be no different as device makers create application-specific flash-based products.

It's not the media that defines the device, but how's it's engineered and augmented to fit application needs.

[Jim Hankins]

Enjoyed reading your view on Enterprise Flash and I agree with you that there is a lot of confusion around this technology right now. Unfortunately, I believe some of it is brought on by some pretty outlandish claims on the market adoption of SSDs, particularly in enterprise disk arrays. For more on our view of this technology, I have written a couple of blogs at: http://www.communities.hp.com/online/blogs/datastorage/default.aspx

Thanks,

Jim Hankins
HP StorageWorks

[the storage anarchist]

Welcome to my blog, Jim!

I appreciate your position that SSDs aren't going to replace or eliminate disks in the next two years. We could debate that, I guess, but to what end? Does it really matter if it happens in 2 or 3 or 5 years - or even never??

As to if/when flash replaces 15K rpm disk drives, I'll offer that they don't have to actually cross-over 15K $/GB to have a significant impact on 15K demand.

As the existence of 15K drives themselves attest, we know that people will pay more $/GB for extra performance. And some will even buy multiple 15K rpm drives and under-utilize them just to get the IOPS they need.

Since a good enterprise flash drive can deliver significantly more IOPS/GB at response times unattainable by spinning rust, using less power and space/IOP to boot, it's not much of a stretch to predict that flash SSD TCO will get "close enough" to impact demand for 15K disk drives.

In fact, it's already started.

While you and Fujitsu's VP of Business Development are arguing about if or when, EMC customers are today displacing 15K rpm disk drive purchases with EFDs for their DMX4's.

BTW - I was surprised that you would choose to quote the opinion of Fujitsu's VP as support for your position on flash. If, as you say, HP has been testing flash "for some time now," then you know that Joel Hagberg's perspective is obviously based on his understanding of the characteristics of MLC-based notebook flash and not of the SLC-based enterprise flash devices that EMC is currently shipping (I'll give you the benefit of the doubt that yHP has had access to the same drives by now so you know the difference between the two).

Your inclusion of his apparently uninformed opinion serves only to promote the misinformation about enterprise flash drives. If you really understand the technology, I don't understand why you'd want to add to the confusion.

John Webster's article really doesn't support your position that Flash isn't happening either. What John said is that customers have to understand what to do with the technology before they can build it into their thinking and plans.

I couldn't agree with John more - customers need clarity.

Fortunately, EMC's customers are already able to factor the technology into their planning and acquisitions.

The reality is that flash drives that can meet the performance and availablility requirements of the enterprise are here today. Whether or not they diplace or eliminate disks is really irrelevant - those that can benefit from their performance will do so, despite all the FUD being promoted by vendors who don't yet offer them in their storage products.