Data Center Tech Blog

Last week the SanDisk® Enterprise Storage Solutions team was at Oracle OpenWorld, and I had the opportunity to appear on the live webcast called The Cube, hosted by SiliconANGLE founder John Furrier, and Wikibon analyst Dave Vellante. We had a great chat about SanDisk’s enterprise business and the future of flash and virtualization.

The full transcript is below. It’s a bit long for a blog post, but if you’d rather read the interview than watch/listen, here it is:

John:
OK, wer’e back live here in San Francisco. This is SiliconANGLE.tv, SiliconANGLE.com. The Cube is our flagship program. We go out to events and abstract the signal from the noise. I’m John Furier, the founder of SiliconANGLE.com and I’m joined by my co-host…

Dave:
I’m Dave Vellante of Wikibon.org. And we’re joined by Rich Petersen, who is the director of software marketing at SanDisk. Welcome to The Cube.

Rich:
Thanks very much.

Dave:
And we were talking off-camera… You were recently acquired by SanDisk. You came from FlashSoft.

Rich:
That’s correct.

Dave:
We’re going to talk about that. John, flash is obviously taking the market by storm. We’ve written about it a lot on Wikibon, and SiliconANGLE…

John:
We’ve covered the Flash Memory Summit. We’ve been deep on flash and solid-state. Dave, this is just a big innovation. Every technology revolution has a new enabler, a destructive enabler and… and spinning disk has not been one of them. So, we’ve been covering that. My question is, tell us what’s going on with you guys now. SanDisk, obviously everyone knows about the brand, anyone that’s ever taken a picture, or seen some of the cool tech that you have, and there’s a variety of other enterprise solutions. So take us through some of what’s not known and what you guys are doing it within SanDisk.

Rich:
That’s a great question. SanDisk is well known as a leader – a pioneering leader – in flash technology. What a lot of people don’t know about SanDisk is that it also has a very active, vibrant enterprise business, oriented around flash technology, both as a developer of hardware solutions and also software solutions that leverage enterprise flash in order to reduce I/O latency. And this reduction in I/O latency is one of the key enablers for increased performance and scalability for databases, enterprise applications, and virtualization environments.

Dave:
So it’s happening pretty fast. It started with consumer devices, and now it’s moving into the enterprise. We are here at Oracle OpenWorld, and Oracle is talking flash. What’s your angle on that? Oracle’s putting flash in a lot of different places. What does that mean to you guys? Obviously that’s good news because it rises the tide, but at the same time Oracle’s getting more vertically integrated with flash. So talk about where you fit.

Rich:
That’s an excellent question, because one thing that’s important to understand is that there many different ways to apply flash technology to the task of reducing I/O latency. In some cases the optimal approach is to approach the front end of the database itself. In other cases implementations of flash in storage architectures provide great benefits. Our particular approach is to enable the flash to act as a cache within the I/O data path of the operating system itself. So whether you’re running an Oracle database on Linux, or something on Windows Server or even in a VM, you can accelerate the performance and scalability of the application by reducing latency through the use of flash technology as a server-attached cache. So that’s where we fit in the stack.

Dave:
Okay, and and the point of management – is that you guys doing that management, you’re relying on the host to do that?

Rich:
Yes, and yes. Essentially the FlashSoft software appears to the operating system as a driver the in the I/O path between server and storage. So essentially, we’re transparent upward to the databases and applications, and we’re also transparent downward to the storage infrastructure. This is extremely valuable to customers that want to deliver the benefits of enterprise flash but they don’t want to re-architect their databases, and they don’t want to re-architect the storage investment that they’ve already made.

Dave:
So you talk in SCSI protocol for example…

Rich:
Sure, we will talk with any underlying storage, and we can attach – or talk to – the flash device via any attach interface, whether it’s SATA, SAS, PCIe even NVMe. We’re very transparent within the operating system, both to the storage protocol as well as to the interface to the SSD.

Dave:
Okay. So where do you see this whole space going? Will all active data sit on flash? Is it going to be more aggressive than that? Will flash replace spinning disk like it has in a lot of these devices? What’s your vision?

Rich:
I’m still waiting for spinning disks to replace tape completely, so I think we’ve got a few more years of spinning disks in the enterprise.

Dave:
Well, it is finally happening. It’s 2012, almost 2013, and it’s finally happening.

Rich:
But truthfully, with the massive growth in data that enterprises are having to handle, spinning disks will be with us for a while. But one of the most important drivers for this adoption of flash is the fact that we need to get access to larger amounts of data much more quickly, whether we’re talking about big data analytics, mobile applications… All of these new applications are placing demands on storage infrastructures that we couldn’t have imagined a few years ago, and flash technology can step up and deliver large volumes of data to those application requirements, with the performance that that we need to do business.

John:
Rich, what has flash done in your opinion to the architecture of the enterprise, within the data center and within cloud in particular, and two, how does that impact of some of the big legacy stuff like big SANs? Obviously, flash can sit in a lot of different places: closer to the processors, on the network… now we have software-defined networking… all these great emerging trends. Obviously flash is this ingredient that’s very cool – it sits in different spots. How does that change the architecture?

Rich:
It’s changing it in a number of interesting ways. I think if you want to look at the development of enterprise flash in the enterprise, there are two ways to look at it. One is the capacity and the performance steadily increasing just in terms of the hardware, but then additionally we should consider the fact that we’re getting better and better at putting flash where delivers the greatest benefit. So initially it was implemented as a tier in a storage system. Following that, we’ve come up with innovative technologies that put it on the SAN. What we’re doing now with technologies like FlashSoft is we’ve found a way to put the flash right next to the CPU, where it can deliver the data fastest, and that’s really changed what servers are going to be looking like over the next several years. We’re going to start thinking of servers in terms of “how large a flash-based cache can I put into this hardware?” and “how can I best leverage it with my existing storage?” One of the things that we emphasize is the ability to leverage existing storage investments because these are systems that aren’t that old, they are very expensive, and there’s a lot of policy and process that’s baked into their implementation. So we want to be as non-disruptive as possible while being as innovative as possible.

John:
So let’s talk about the customer problems that are being solved. There are problems to solve, business to be done, and as you mentioned, a lot of legacy investments. So what are the core problems that they’re solving? Are there some things that bubble to the top in almost every conversation?

Rich:
Oh, absolutely. I would say at the top of the heap is is virtualization. Today, most organizations have picked the low-hanging fruit in virtualization, and they’re moving with a roadmap to increase their virtualization, and yet they’re still looking at certain tier-one systems that they’re reluctant to consider virtualizing. Flash technology enables them to expand the scope of their thinking and look to virtualization in areas where previously they might not have imagined doing so. Secondarily, is what we see people doing in terms of databases, where they’re working with very, very large amounts of data. They’re looking at business analytics, they’re looking at OLTP implementations, where they’re really increasing their expectations for performance. They’re looking at batch jobs that can be brought down from thirty hours to under ten hours, sometimes five or six hours. They’re looking at response times dropping from seconds to milliseconds. That’s how flash is raising expectations for what applications can do within the enterprise today.

John:
Tell the folks out there: why SanDisk? There’s a lot of noise in the marketplace, a lot of different vendors. In the enterprise, SanDisk is not as well-known, relative to the awareness for its consumer brand, so why SanDisk? What are some of the reasons why people buy you guys and work with you?

Rich:
SanDisk is actually a very partner-oriented organization. We don’t have a dog in many of these fights. For example, to talk about the FlashSoft technology, even though SanDisk is a manufacturer of enterprise solid-state devices, our software technology is open to be used with solid-state technology from other vendors, whether you’re talking about LSI, Fusion-io or Virident. So it’s one way that you can see that as a large organization with a diversified portfolio, we’re able to partner with vendors of storage systems, partner with vendors of servers, and partner with other SSD team makers as well. We can say to the customer, “you truly are our top priority, and your procurement decisions are your procurement decisions.” We will support that, whether we’re talking about our hardware offerings or our software offerings.

John:
What’s the technology status these days? Obviously flash has continued to grow, price points are going to drop, in a continued evolution. Flash is still kind of expensive relative to… well it’s expensive, but in the scheme of things it can pay for itself depending on how you implement the architecture, which we have talked about in The Cube many times. But what’s next in the technology? What’s around the corner that’s going to continuously improve the capability?

Rich:
Obviously, capacities and performance characteristics are going to change. One of the things that I think a lot of people almost overlooked is the degree to which the longevity, the consistency, the reliability of these devices has improved radically. And this is an area where SanDisk has been a pioneering vendor on the hardware side. Really there’s a lot of under-the-surface improvements in enterprise flash technology that maybe don’t grab headlines in terms of “millions of IOPs,” but they really grab attention when the CIO is looking at the long-term value of his investment.

Dave:
I wonder if you could talk about the landscape a little bit. I think that the number of suppliers in the space is just exploding. You guys have been here from the beginning, you’re probably the largest supplier through the value chain. You’re a public company, very successful, with revenues in the multi-billions, and a market cap of ten billion. But you see a plethora of companies, many are very niche-oriented, focused solely on the enterprise. But you are a very diversified business. You’re in consumer, you’re in the iPhone, you’re in the enterprise… and FlashSoft brings another capability. What’s the overall strategy of the company, relative to these emerging companies that you see with great valuations, that are getting bought sometimes before they even have a product?

Rich:
SanDisk is a company that has a great deal of experience and a great deal of expertise in-house. So, for example, FlashSoft became a part of SanDisk in that very process. We were shipping a product with a good customer base, we had demonstrated some leadership in this technology area, and that’s where the folks at SanDisk were able to execute on an acquisition strategy that made sense for them and made sense for us as an organization, and in particular made sense for our customers. The number of business partners and customers of ours that said “we’re really glad to see you become a part of that organization because that’s somebody we can do business with…” They saw the depth of technical expertise, they saw the openness in terms of business strategy and they saw the leadership at the corporate level. So it really gave us a boost in terms of where we saw this technology going over the long term.

Dave:
A lot of people talk about something else on the horizon other than flash. You know, they say that if we had to choose a technology for the enterprise, we wouldn’t have chosen flash, because you have to go through so many gymnastics to make sure that you don’t don’t lose data. At the same time, the compelling part about NAND is that it’s in all these consumer devices and the price dropped dramatically. What is SanDisk’s outlook on that from your point of view?

Rich:
Frankly, having non-volatile memory that can be resident in the server is a very important innovation, because it really gets storage capabilities ahead of the latency that’s inherent in any storage area network. By bringing storage to the server with non-volatile memory technology, I think we’ve pioneered the first step in a new kind of architecture, an architecture that is better for virtual machine environments, it’s better for large-scale database implementations and I think we’re going to see both server and storage vendors aggressively supporting this innovation. So I’m a very strong proponent of non-volatile, flash memory in the server. I think that’s really the future.

Dave:
I think we would generally agree with that. At least for all active data, it just makes sense to put it on flash. And the cost from a transaction standpoint is actually lower than on spinning disk. But what about flash going away in favor of some future technology like memresistors or something like that? Can those technologies compete with the massive volumes of the consumer business, or is flash here to stay for the foreseeable future?

Rich:
Well, I’m going to beg the question, I’m afraid, because I’m a software guy.

Dave:
So you don’t care.

Rich:
In a way. But there’s always going to be something faster. There’s always going to be another tier that’s faster than the previous tier, and there’s always going to be a requirement for the intelligence to understand where the data is best served. From our perspective that’s just a very bright future.

Dave:
So to follow up with a software question, it seems like a big software innovation is being able to actually access and act upon the metadata.

Rich:
That’s right.

Dave:
Talk about that a little bit. Who controls that metadata and how does that shift and what does that mean for your business?

Rich:
That’s that’s an important question and it’s one of the issues that again, when I talk about innovations that are below the surface – that’s one of them. Because metadata management is really something that, if done poorly, can overload systems so you’re basically going to be stealing from Peter to pay Paul. In order to accelerate your storage I/O, you may be stealing resources away from system memory and in-memory caching capabilities.

Dave:
Pushing the bottleneck around.

Rich:
Exactly. So it’s extremely important when you’re looking at a storage technology, you have to make sure that metadata management is implemented appropriately. That’s one of the things that the data scientists on the FlashSoft team have made a particular point of focus. Our footprint, if you will, in the server is extremely light. If you don’t mind some data points…

Dave:
Please.

Rich:
In our current VMware product, the maximum size of the SSD-based cache can be 2TB, and that 2TB cache can be managed with 140MB of system memory. So it’s a very light metadata footprint in order to manage an extremely large cache of data. There are some special architectural developments that we put in place and order to be able to do that, but I think when you look ahead into the future, as data grows obviously metadata is going to grow as well. And the question is, are we developing technologies that implement metadata management for maximum efficiency? Because that’s really one of the subtle differences that will come to the fore when you’re talking about the pros and cons of different approaches to solving this issue of I/O latency.

Dave:
Well if you get it right, the performance impact is going to be enormous. Talking about VMware, recently VMware put Pat Gelsinger in charge. He brings a hardware background – a systems background really. He has a lot of experience with his former company EMC. David Floyer, whom you know, wrote that where they’re probably headed is basically to try to control the entire stack in the metadata in particular, saying “OK, we own the protocols you gotta play with us.” How do you see that as a software supplier? Where do you compete long-term? Do you just say “OK, we will participate in whatever protocols we have to interact with, and add value through our efficiency in our algorithms”? What’s the long-term prognosis for you guys in that regard?

Rich:
That’s that’s a difficult question to answer directly, because I don’t have a crystal ball. Our relationship with VMware has been a very strong one. We’ve been a VMware partner from the beginning, and one of the things that has has impressed us the most is when you look at VMware’s partner ecosystem, it is very robust. And I think it’s very robust not by accident, but by a dedicated effort to support its partners and to involve them in the innovations that it brings to to its customers. I obviously can’t comment or speculate in any way as to what decisions Pat Gelsinger may make, but as a company I think VMware got to the to the point where it is through strong partnerships with storage vendors, server vendors, software developers; and encouraging innovation in its ecosystem. So I’m not skeptical – I’m much more optimistic.

Dave:
So said another way, you don’t see VMware or EMC – its owner – sub-optimizing the performance of its long-term vision to maybe prop up VFCache or something like that; rather you think the ecosystem will generate a lot more value for obviously the industry but also VMware itself.

Rich:
Very much so. I think VFCache brought to fore the requirement for server-resident, non-volatile memory, specifically for the purpose of improving the performance and scalability of virtualization environments. So when a large vendor like EMC brings a technology like that to its customers and says “this is how we see the future,” to some degree that raises the awareness among the customers of EMC’s products – of all products – to the importance of bringing non-volatile memory into the server. Prior to that announcement there were a great number of skeptics about bringing what was considered a storage technology to the server tier. So to some degree it’s done us a service.

John:
So we’re getting close on time, but I want you to comment on what we have been following as the hottest trend since VMware bought Nicira: software-defined networking, or network virtualization, or software-led infrastructure, or software-defined datacenter – whatever name the marketing guys are going to try to put on this mega-trend – it’s pretty relevant. We’re talking about software now at the network layer, let loose virtualization magic in there… This is only going to help proliferate more cache. So what’s your take on this? One, the direction, the state of the technology, and its path?

Rich:
If I could speculate slightly…

John:
Go ahead.

Rich:
I would say that one of the things to keep in mind is that caching technology observes the I/O behavior in a server and storage system. And in doing so we always think of, well what does it do with that information? Well, it simply uses that information to assign hot data to the appropriate tier of the stack, if you will. But what else might we do with that information, in terms of optimizing the way our data centers operate?

John:
Routes? State information?

Rich:
Exactly, there are many different things we might do.

John:
There are opportunities all over the place.

Rich:
There are incredible opportunities, because I think over the long term, this flash-based caching technology is thought of only in terms of what can it do for performance and scalability, but keep in mind it is generating a lot of information that we may be able to do something useful with.

John:
You mention the word “policy.” People think about political policy, and policy-based networking and policy-based storage – everything’s policy-based that requires decision making with dynamic information. Software is the case. We are obviously bullish on software-defined networking, the software-defined datacenter, all the above. So I agree. I think it’s a great opportunity. Thanks for coming on The Cube. I really appreciate it.

This is Oracle OpenWorld. We are here live here from The Cube, our flagship program, abstracting the signal from the noise. Rich, thanks so much. We’ll be right back with our next guest after this short break.

Rich

 

SanDisk® CEO Sanjay Mehrotra on FOX Business' Three Days in the Valley

¡Hola! from VMworld 2012 in Barcelona

Subscribe Today!

Get our latest posts via email and enjoy data center insights in a flash