This content is part of the Conference Coverage: EMC World 2016: One-stop shop for conference coverage
News Stay informed about the latest enterprise technology news and product updates.

EMC flash storage discussed by distinguished engineer Daniel Cobb

The introduction of a new Unity midrange system to complement the company's XtremIO, VNX and VMAX arrays is another piece in the evolving EMC flash storage strategy.

During an interview at EMC World 2016, Daniel Cobb, a vice president and distinguished engineer at EMC, admitted there may be overlap among the all-flash versions of the products, but he said the company is comfortable with it. Once Dell's acquisition of EMC closes, customers will have even more choices of all-flash arrays.

Cobb discussed the impetus for the Unity product, the demand EMC saw for a new midrange array among the company's customers and the key differentiator for this particular EMC flash storage product. He also addressed questions about the future of EMC's VNX arrays, as reports swirl that the product might be discontinued.

"We're going to continue to have VNX in the market. There's a very loyal and loud customer base for VNX," he said.

But, Cobb added, "people who love VNX are going to love Unity."

EMC is banking on triple-level cell (TLC) 3D NAND flash with the new Unity all-flash array. Cobb noted that EMC also uses TLC 3D NAND with its VMAX platform announced in February. He said TLC 3D NAND flash is now ready for prime time, even for high-transaction databases.

Cobb said the ability to "move backwards in lithography" is one factor that has worked in favor of the EMC flash storage approach. "The care that you had to take when you were managing, say, a 15 nm flash device is not the same [care] that you have to take when you're managing NAND in a 3D device. It's a much friendlier environment in terms of endurance, writes, error rates and things like that," he said.

"We've [also] done the changes in our system software that allow us to be much friendlier to the flash in general," Cobb added. "So we get commensurate benefits that come from having software that treats flash in a way that gets the most out of it."

When explaining how EMC flash storage will be managed, Cobb said, "because we're going to larger capacity drives, we have much more flash in the system essentially to write to, and so we don't end up with hot spots even at an individual drive level because we're managing flash entirely across the complete system," he explained. "That allows us to spread those writes out amongst a much larger population of flash. And we just don't have the hot spots anymore that we had to manage when a system might only have contained 1%, 2% or 3% flash."

View All Videos

Transcript - EMC flash storage discussed by distinguished engineer Daniel Cobb

Sliwa: Hi, I'm Carol Sliwa. I'm a senior writer with TechTarget Storage Sites. I'm here at EMC World 2016 with Daniel Cobb, a vice president and distinguished engineer at EMC. First of all, I'd like to know what the impetus was for the new Unity Flash product line given that EMC already has several flash offerings and has more on the way from Dell.
Cobb: One of the things that we've been looking at for a long time is how we [can] accelerate the industry transition from hard drive technology to flash technology. There are a lot of benefits that accrue when you can do that. When we first shipped flash, as in hybrid storage platforms, we acquired XtremIO to have a net new platform for all flash arrays. We now sell flash VMAX in February and DSSD as a more server-centric or cloud-native platform, three-centric platform, for that.
We didn't yet have that bread-and-butter dual controller midrange architecture covered with our VNX products, so with Unity, particularly all-flash Unity, what we're able to do is get in under a $20,000 price point with a very useful flash footprint for remote offices, branch offices, or those smaller enterprises where a $20,000 entry point in an all-flash configuration is something we never would have hit with any of our other platforms.
With those four platforms we feel like we've covered the industry, we've covered the use cases in the market with the most comprehensive portfolio that exists in the flash space today. We're comfortable if we have a few overlaps between platforms. We've always been comfortable with that, and we'd rather do that than leave a gap for a competitor to sneak into. So we really like our position with our four platforms right now.
Sliwa: What happens with your VNX platform? Does that continue?
Cobb: We're going to continue to have VNX in the market. There's a very loyal and loud customer base for VNX. There are people who go all the way back to the Clariion days and Data General. People who like VNX are going to love Unity. You know, essentially, what we're able to do is leverage a lot of the data service maturity that people got used to in VNX. We essentially jacked the house up and put it down on a new foundation.
So with Unity, we have a foundation that's flash optimized, that's multi-core optimized, that has a new RAID and cache implementation, and things like that. But a lot of the other software stack capabilities, we were able to take from VNX, so we've still got that mature reliable enterprise-grade set of services that those customers value.
Sliwa: What about Dell's all-flash products? They have all-flash versions of the former Compellent and EqualLogic arrays.
Cobb: To a large extent, the product lines are very complementary. Obviously, there are' parts of the Dell integration situation that I can't speak to, but if you look at what's available from IDC and their price bands analysis and things like that, it's astonishing to watch how complementary each of the 'offerings are across the various price points. From low-end RBOD-type write controller things where Dell's been very successful in their PowerVault line, moving up through into the areas where there is some overlap in the midrange. We'll have to sort through that, but there is much less overlap than is apparent at face value.
Sliwa: What advice do you have for customers who have to sort through all these different all-flash product lines from EMC and from Dell?
Cobb: Well, I'll answer for EMC. I think the principles are large enough that they'll apply whatever the portfolio, but we look at it in a fairly straightforward way. 'There's only one platform on the planet that can manage the devices and have thousands of drives, thousands of ports. There's only one platform on the planet that can do that, and that's VMAX. If you need a consolidation platform at that scale with that variety of data services, synchronous and asynchronous replication, that's been field tested for 25 years now, that platform is VMAX. It's the gold standard. That's why all-flash VMAX exists.
If you're looking for a platform that was designed specifically for flash, including always-on, always-inline data services, deduplication, compression, copy management, all of those things done in a totally integrated, native flash way, that's XtremIO. For some people, that's been their first EMC platform. They've looked at some of the other products, but when they saw XtremIO, they said, 'We're ready to consider EMC again.' And that's now the largest selling flash platform on the market.
DSSD is our top fuel dragster. If you're looking for speed, if you're looking to run as fast as the hardware can natively run without a lot of software in the way but with the enterprise-class robustness and reliability of an external storage platform, that's DSSD for those cloud-native, low-latency, high-frequency applications.
And then the bread-and-butter midrange platforms start to come in a little bit lower. You can start at $20,000. They can be your remote office or branch office footprint. They can be your departmental vSphere footprint, shared storage footprint, and things like that. And now you can get into that platform, all-flash, for less than $20,000.
So, we're giving people some pretty distinct choices to make and allowing them a choice of four platforms that really run the gamut of all the use cases we see in the enterprise.
Sliwa: Is there anything truly unique about your new Unity product line?
Cobb: With Unity, the focus there has been primarily on simplicity. Customers in midrange and lower end shops don't want to spend a lot of time on data center operations. They don't want to spend a lot of time setting up a storage platform. So, simplicity out of the box is just set up to minimize clicks, minimize intervention, minimize Opex costs that are associated with storage in those [types] of low-touch IT shops. So when you build a platform that's optimized around that, you take a different approach to how you might manage thousands of ports connected to a VMAX and hundreds of applications running simultaneously at scale. It's a different problem and it merited a different solution.
Sliwa: Did you have any sense there was real demand for the Unity product line from your customer base?
Cobb: Yeah, we really did. In fact, you know, we've been watching for a long time, and essentially we've invested significant dollars in helping the flash industry. We've been watching this cost decline of flash. It's driven, fundamentally, by Moore's Law, but it's also driven by the ability to go to 3D and flash. It's been driven by the ability to not store just one or two bits per cell in flash, but three bits per cell. So, now that the industry has three-bit per cell 3D flash, we were able to shift our entire hard drive strategy in VNX over to flash in Unity because of that enabling technology. So that really allowed us to do some net new things with that platform that we would have had a harder time doing with some of the legacy that existed before.
Sliwa: Are you using 3D NAND in the Unity product line?
Cobb: Yes, we are.
Sliwa: So you're using TLC 3D NAND with the Unity product?
Cobb: [We have] TLC 3D.
Sliwa: TLC 3D NAND hasn't always been the technology of choice for heavy enterprise workloads. What are the pros and cons that you see for TLC 3D NAND, and how did you overcome the cons?
Cobb: The thing to keep in mind with flash, we started with single-level cell flash and for a while, that was the only thing that could ever be used for enterprise workloads because it had the best performance and the longest endurance. It didn't wear out as often and for the types of high-frequency workloads that you would see in an EMC platform. That made a lot of sense. It made customers feel a lot better as they considered a new media technology. They were comfortable with that and then it was our turn to get the industry comfortable with MLC flash. We had invested significantly in MLC flash controller companies out there and we helped the industry transition from SLC to MLC.

Along comes 3D flash and for the first couple of generations, it still wasn't quite more cost-effective than 2D flash. The initial 3D stuff was a little bit more expensive. The second generation came out on par, but that wasn't a significant enough push to tip the industry over. The third generation 'is available now, and we see that available from Samsung, Toshiba, Intel, Micron, SanDisk, et cetera. We see that as very enterprise-capable because when they moved to 3D, they were also able to move backwards a little bit in terms of lithography.
So, the bad things that happened when the semiconductor industry was shrinking flash in two dimensions; higher error rates, slower write times, things like that, those went away when the industry was able to move backwards, and moving backwards in lithography but then stacking in 3D got you to capacity. Moving backwards in lithography also let you get to TLC because you had, essentially, a friendlier media to be able to program, and the friendly media is what enabled TLC.
Sliwa: Are there any enterprise workloads you would not put on TLC 3D NAND in the Unity product line?
Cobb: No, I think we're ready with our Xpect More Program to essentially warrant flash devices or whatever devices we ship for the workloads that customers demand for the lifetime of the array, and we're through the concerns where we had to tell people, 'Watch out for this, don't do that.' Our systems software is good enough now, the way we manage flash drives in systems, takes all those things into account and now, essentially, we worry about those things so the customer doesn't have to.
Sliwa: So are you saying TLC 3D NAND is ready for high-transaction databases?
Cobb: Absolutely, and that's why we're here saying 2016 is the year of all-flash for production.
Sliwa: Are you using TLC 3D NAND in any products other than Unity?
Cobb: We're using that in VMAX as well. 

+ Show Transcript