 Live in Silicon Valley in Santa Clara, California at the convention center. I'm John Furrier, the founder of Silicon Angle. This is theCUBE, our flagship program. We go out to the events, extract the signal from the noise. Enjoy my co-host, Jeff Kelly from wikibond.org. Big Data Analyst here live at Percona Live, where all the access is happening around data, databases, all the future, under the hood of cloud mobile and social and big data. We're here with Gary Orenstein, Cube alumni, did very well in our Cube madness bracket for a while there. Gary, welcome back to theCUBE. Thank you very much. Thanks for having me. So are you president of EVP? What's your title now? You seem to get- I'm the marketing officer of Fusion IOS. Okay, so that's the focus, yes. I couldn't help but get that joke in there, but it was good. You've been moving up the ranks in Fusion IOS, a really important company around web scale. And a lot of stuff happening here that a lot of people in the mainstream press have written about some web but mostly in the geek circles and that's the web scale sequel trend. Yep. And on the opening, Jeff and I were talking about the Cloudera news monster, four billion dollar valuation. Ridiculous, some say bubble, some say innovation. But what we're commenting about is data first is seems to be where the cutting edge action is. Not cloud first, not mobile first, data first, data as a new oil, new currency, but from the development standpoint, it's a real innovator. So I want to get your perspective of this data first environment and I want to jump in and get your perspective on this web scale, my sequel stuff, my web scale sequel that Facebook, LinkedIn, Twitter, and Google are rolling out. Absolutely. So there's no question that with what we see today, data is becoming more and more important. I saw a report yesterday about the shortage of people who know how to analyze data and work with data and how that's going to be a real gap as companies try to further their business. But we're thrilled to be participating at Percona Live, the biggest show of the year around the MySQL activity and wanted to also share some of the news that we brought to Percona Live today. We announced some new features around what we can call flash-aware applications. So Fusion IO is very focused on delivering flash solutions to both hyperscale companies and enterprises. And we've always been searching for ways that we can optimize the whole data stack, not just optimizing at the storage layer, but thinking about what's the path from where the data resides to the database, to the application, all the way out to the end user. And so in doing that, we've come up with some optimizations around flash memory. And the one that we announced today is called NVM compression, non-volatile memory compression. And what that allows people to do is in compressible workloads, get up to 2x the usable space for their data. So that's a huge benefit. And the other part that we've introduced with NVM compression is that there's no overhead now for using compression. So compression's been part of MySQL for ages and ages, but it's caused a little bit of an impact when enabled. And so people have been hesitant to use it quite rapidly. We've eliminated that penalty. So this idea of taking flash memory, looking at it in the overall food chain about how do we optimize the data stack? How do we help people process more transactions, use fewer servers, use less power, be more efficient in their data usage? We're seeing examples of that really come to life with the introduction of making flash applications aware. And we're really thrilled that we've been able to cooperate with Percona, as well as MariaDB, as well as the MySQL team at Oracle, who mentioned us in the presentation this morning, to make some of this, bring some life back to this, which we've seen in many areas. And of course, just from the news and the activity here in the conference today, you can tell that there's still plenty of life left in this MySQL arena. So what's happened in this marketplace? You guys have been very successful and continue to be successful in the flash now and with non-volatile memory compression. People want more speeds and feeds, obviously that's one key element, but there's a bigger picture than that. So give us an overview of what's going on of the hyperscale market. Obviously with web scale, you're starting to see some more mainstream penetration around hyperscale. What does that mean for architectures, cloud computing? You heard Satya Nutella at Microsoft saying cloud first, some say mobile first, we're saying data first. A lot of stuff going on in the hood. But what's happening in the market? Give us an overview of what you're seeing. Sure, so just to focus a little bit on the MySQL community in regards to that question, we're seeing across our customer base tremendous usage of MySQL Percona MariaDB to solve web application challenges. And that could be anything from online games to messaging applications, which of course we know are all the rage to online text message voting applications to you name it. We've even seen this MySQL being used very widely in areas for analytics, advertising analytics and consumer insight. And across the board, what we found is that people really like the familiarity and the relational nature of MySQL, but have had challenges where they've needed to perhaps scale out a little bit more than they wanted to or a little bit more than their data center could handle. And so by using flash memory in these instances, we've been able to help our customers shrink their footprint, consolidate the number of servers, get much better efficiency in terms of the number of transactions that they're able to deliver for a given unit of server space or power. And again, across the industry, we're seeing this helping people essentially build applications that couldn't be built at a certain economic number previously. So it was a couple of months ago at the Open Compute Conference. There was a fireside chat with Mark Andreessen and Andy Bechtelscheim. And they were saying that a few years ago, we didn't have the capability to deliver so many bits of information to so many people, so cost-effectively. But now with technologies like MySQL, coupling that in places with flash memory, we're really able to do that. Gary, I wonder if you could talk a little bit more about the web-scale SQL, I guess consortium group, Facebook, LinkedIn, Google, and Twitter getting together. When we think of those companies, we think oftentimes we think about things like Hadoop and some of the innovations they're doing there, some of the new SQL innovations they're doing. But all of those companies getting together to actually help scale and continue using MySQL as the front end for a number of their applications. Talk a little bit about the implications of that. Yeah, so this was an interesting announcement that I saw last week where a group of companies, the companies that were listed were Google, Facebook, LinkedIn, and Twitter, announced this initiative they called Web Scale SQL. I'm pretty sure they have a webpage that points to a GitHub page for that. And they describe this as a branch of MySQL that will be optimized for the type of web transactions that they're working on. And they also announced that they would do this off of a core community branch from MySQL itself. The interesting part here is that we're seeing how important MySQL and MySQL-like databases are to this community. And one of the interesting things about that announcement to me is when I think of those companies, Google, Facebook, LinkedIn, Twitter, I think about companies that I do trust with keeping my data, allowing me to interact with my data, really providing a very rich infrastructure in order to do that. And I think it's no coincidence that that group got together and said, the place for us to start with this is MySQL. We need to do some optimizations on top of it that are suited to our workload. But they didn't go off and build something net new and they didn't go off and just grab the latest shiny new toy. They went to a very rich, robust platform in MySQL and said, we're gonna tinker with that to optimize it. And so that's my take on it. And one of the things that's interesting that, again, that premise that you have in interacting with those companies is one where you're really counting on them to keep your tweets, keep your photos, keep your Google documents, keep your profile and your professional history, and they chose the right tool for that. Gary, I want to ask you about the MySQL market. Obviously people in the business know it's obviously a huge open source program. Obviously Oracle owns it now. But you look at some of the numbers in about a quarter of the multi-client DBMS as database management systems run on MySQL. It's been a great open source. LAMPstack developers have been embracing it. And you saw companies like Facebook build their own memcaches, and memcache was then open sourced, and you got Amazon with elastic care, all this stuff around under the hood kind of things that are required for scale becomes huge. So now with the web scale SQL model, you can now bring a path for hundreds of thousands of developers to go massive scale. So I get that. So I got to ask you, is it unique that these companies are using data at the heart of their value proposition? And is it a corner case? I mean, we all know Facebook uses the data, Google uses the data, and they have privacy issues in their hardcore about that. But for the average enterprise, Mongo's had huge success, but their ding on Mongo is scale. So there's a glass ceiling, if you will, of performance. What is the nuance between massive scale and what MySQL can do up to now, and how much will enterprises really start using the data? Or is it they just scratching the surface right now? Yeah, there's obviously, there's many ways to tackle these problems. One of the things about MySQL that is proven is there's very successful models for sharding or partitioning the database to grow it. And certainly companies like Facebook have shown that, but certainly many others as well. And so this whole debate about scale is somewhat, it's ironic in that there are ways to scale MySQL to support the largest bits of web infrastructure in the world. It's a different scaling model, yet it works very well for what it does. And I do think that whether it's with MySQL or another data store, that companies are becoming increasingly reliant on their ability to capture, process and present information. That sequence of, you want to both be able to ingest what you can ingest, you have to be able to run the analytics and run the queries and do so without taking your infrastructure down and then deliver that in some type of actionable insight. Those are the things that I find we see with our customers who are struggling to do this. Obviously, adding flash to the equations, you can have flash infused data to speed up those workflows, particularly to handle that kind of multiple workflow type of process and how to both capture the data, process the data and present the data. That's becoming a critical skill. And whether you're in the web scale, hyper scale world, or whether you're in the enterprise world, I think it's pretty widely seen that companies of the future are going to master that skill as one of their core operations, day to day operations. Gary, let's talk a little bit about this show in particular. We were talking before we went on a little bit about the mix of people here at the show. You've got some vendors, of course, you've got a lot of participants in the, or I should say practitioners in the community as well who are here, such as Facebook. What does that say about this show and about this community and the state of the community around MySQL right now? Sure, well, I'll say I'm fairly recent to this community, been involved the last few years, but I have to say it's great to see a venue such as Percona Live with a mix of both industry participants, but also customers. We see Facebook here, we see Dropbox here, we see Yelp here, I'm sure there's several others that I didn't get a chance to visit yet. And what a wonderful opportunity to have this level of engagement between both the development side of the house and vendors that are providing different solutions to help solve these customer challenges, but actually seeing the customers here presenting on their own. I think it's unique to the trade show industry and technology and something that I think showcases the strength of what MySQL built and what companies like Percona, like Oracle, like MariaDB are carrying on. Gary, talk about the non-volatile memory compression because there's been a lot of discussion in the geek circles now hitting mainstream around what Flash has done for the programming models. Obviously the Linux kernel was written in the day when you had a lot of storage of non-volatile memory. Now you have a lot of it and this seems to be the bottleneck now. So what does this compression mean for current customers, current environments? What is the thrust of that initiative? Yeah, so let's take a little bit of a step back and then we'll address that specifically, which is when you look at that data stack or that transaction stack of how do I go from the piece of data that relies on some kind of storage media out to the end user. In between there you've got some kind of storage infrastructure, you've got a server infrastructure, you've got operating system, database application, and then end user on their new fangled mobile device. Ever since FusionIO began, we took a different tack. We said as this new memory was coming out, this new medium of flash memory, how can we use this to the best advantage and how can we be different from the architectures that may be tying us to the past? One thing that we did is we said, let's not use storage protocols on this new memory, let's expose it natively to the CPU and the operating system so that it can operate quickly as opposed to operating using protocols that were designed for rotating spinning media. That philosophy, that design approach of thinking about a new memory in terms of its potential for the future as opposed to just thinking about it as how it would fit with the past, we wanted to do both. Since then, over the course of the last couple of years, we've been able to look hard at the applications and the databases that rely on storage, previously relying a lot on disk drives and DRAM, now flash memory bringing this great new memory tier that is a little bit in between. Flash memory is persistent, which is similar to disk drives, but it's also really, really fast, which is a lot more similar to DRAM. And as we've gone in to explore, what we found is that in addition to just allowing the flash memory to act as a fast data store or act like a fast piece of block storage or act like a fast bit of disk, there's much more opportunity on top of that with just minor interaction with the database or the application. So in the case of compression, we looked at MySQL compression and in our interviews with the community, we found that it wasn't used very widely and you'd say, well, gee, if I can compress my data and get more usable space, that's a no-brainer. Why wouldn't I do that? Well, it turns out that there was some overhead that you had to pay a slight performance penalty in order to use the compression and so a lot of people just didn't go there. What we've been able to do by exposing some of the APIs and primitives from flash memory to MySQL and some minor tweaks to MySQL, which we've worked with the community, we've worked with Percona, MariaDB, the MySQL team at Oracle, is we've been able to essentially eliminate that overhead impact. So you can now utilize the compression that has been built into MySQL in a way that doesn't provide you any performance impact, it's essentially now giving you everything that you wanted before plus a big gain on your usable capacity. So what is it? That's just going to drive cost down. So again, the idea here is we're not satisfied just by saying we're going to deliver a great flash memory product. What we're out to do is say, how are we going to optimize this whole infrastructure so that we can allow our customers to deliver more transactions at a lower cost? And I'll take a slight transgression here because one of the challenges people ask, well, what's taking so long and how come this stuff isn't moving faster and what's going on? We're really always trying to work with customers to evaluate the benefits of these kind of technologies in a calculation that makes sense for their application. So to be specific there, since flash memory is a storage product, a lot of people say, I'll evaluate that in terms of dollars per gigabyte. And that's okay, but flash memory is much more about serving data than it is just about storing data. And so we move to talk about, let's look at a dollars per IOP perspective and we're getting a little bit better there as now we're looking at how we're serving this data worldwide. You can go even further and say, let's look at it in terms of a dollars per transaction perspective, how do we actually optimize that? And the culmination of the story is that late last year, December, eBay shared some information about their knowledge, sorry, their key performance indicators. eBay is measuring URL served per kilowatt. What are interesting idea that they've understood that as a company, they're serving web pages in a sense for the auction. We know how much we all refresh when we're doing an eBay auction. They're serving web pages is the most important thing they do and their most cost, biggest cost component to that equation is electricity. So they're optimizing for URLs per kilowatt. So we spend a lot of time working with customers to explain these different capabilities, whether it's NVM compression. We've also done things with atomic rights. We can talk about that later to basically get customers to a better value equation for serving their users. So you're basically bringing out new KPIs for companies. So eBay is a great example. It's kind of nonlinear thinking when you say, hey, energy is a big part of the equation of cost of ownership. Obviously data center we all know what's going on in the data center relative to energy and footprint, you know, AKA got cloud on the horizon and on-premise, which by the way, I went on again for the night, zillion time, on-premise isn't going anywhere in the future. It's going to change, but the idea that on-premise is dead is crazy. But it's on-premise, so it doesn't go anywhere. Exactly. Well, cloud is supposed to kill on-premise data centers, but that's not going to happen. But what are some of the other KPIs, key performance indicators that you're seeing that companies are doing relative to the new, this new modern era of computing? Yeah, so we try to take a very workload specific view and we want to look both at the top line performance and top line performance, we want to get as close to the end user in terms of that performance as possible. So one thing we're seeing now is a big focus on latency consistency. So people always want to have peak performance, but what we're finding is that users are becoming increasingly aware, and our customers are becoming increasingly aware of predictability in performance. And last year we did a customer survey, it was the first time that consistency of performance actually eaked out a little bit ahead of peak performance. And this is an area where Fusion IO is focused a lot in the past. We believe that flash memory and in particular Fusion IO flash memory can provide that kind of predictability that our customers and our customers, customers seek from the experiences that they have with their applications. There is no one performance indicator that fits for all companies, but I share the eBay story and their key performance indicator dashboard is online that companies need to seek out what works for them. And simply looking at the lower levels of the infrastructure, whether it's dollars per gigabyte or dollars per IOP is rarely enough to get and capture the value. You have to look both at the top line performance, the dollar cost, the space cost, the power cost, which are all big contributors here, and come up with the right equation that works for that particular workload. That's an interesting point because I think it also relates to what we were talking about earlier with the whole SQL versus SQL debate. And it's all about mapping, it sounds like to me what you're saying, it's all about mapping the business requirement to the technology. In some cases, flash is gonna be appropriate relative to the business requirement. Some of the other types of storage might be. Absolutely. And same thing when it comes to no SQL versus SQL, there's no one size fits all when it comes to this kind of technology. Yeah, I mean, people love having these debates because it generates a lot of headlines and a lot of emotion. But I think that even here, we're seeing quite a variety of different types of solutions, many that relate to the core MySQL, but not certainly limited to that. And I think it would be foolish to think that we're gonna solve all of the world's data challenges with a single type of data store. So even little variations like we talked about, whether it's enhancing with flash for NVM compression, whether it's enhancing for a web scale workload with web scale SQL, whether it's choosing some net new data store because that's the way of the future. It's just like the online tools that we have today. There's no one tool that's going to solve all of our workflow needs. And given the accessibility that people have to these new types of data stores, given the open source nature of a lot of this work, we're going to live in a heterogeneous world for as far as I can see. Gary, let's talk about the customer environment. So honestly, you know, with, you know, web scale SQL and the things you mentioned around some of the things that eBay's doing and others and innovation, not everyone can be that scale, right? Facebook and Google are great examples. I mean, people say, oh, I want to be the Facebook. I want my enterprise IT to be like Facebook. Good luck with that, right? I mean, Facebook got there from a different route. You know, they didn't have a lot of dough. They self-funded, they built their own gear on open source. You know, they had a lot of manpower. That is not the IT environment of the past decade or so. Now, you're seeing IT start to bulk up a little bit, doing some transformation activity. That's well documented. SiliconANGLE will keep on. But not everyone can ever get to be like Facebook and Google and LinkedIn and Twitter. So what do you guys see there? I mean, can you explain that to the folks out there? Share your perspective of how do companies become like Facebook and Google and Twitter without actually having to have all that muscle? Sure. So the good news is that the, while the scale of those problems might be different between one of these web giants and a smaller company, the nature of the problems is not that different. And jumping back, you know, a couple of months ago, we were at the open compute conference and we spent a lot of time working with partners there, like Cinex and Hive Solutions as well as Quanta because some people found open compute was just a little hard to digest and a little hard to understand. So we worked hard with those partners to have open compute solutions that use flash memory that they can buy in an easy button. You want to buy a rack of servers because that's all you need. You could call Hive Solutions and just do that and they'll deliver it to your doorstep. You could load it up with MySQL and you'll be off to the races. We see a lot of customers who are operating at smaller scale. Maybe they have a dozen servers. Maybe they have half a dozen servers. They're still able to get huge benefits in consolidating when it comes to the deployment of flash memory. So it is not uncommon for us to see customers who may have had 10 servers based on a disk infrastructure previously doing MySQL transactions, move to a flash infrastructure and cut that down to two or, you know, six servers and cut that down to two We've seen extreme cases where people had been scaling out just because they needed the extra CPU and DRAM capacity as much as 16 servers for an online voting application where they were able to scale down to one server. So the same principles are there, which is how do I find the optimum balance between the disk storage I need, the DRAM that I need, the flash memory that I need to create smaller infrastructure that can serve more transactions. And the good news is these principles that we hear from these web giants can be applied at smaller scale and wonderful that we live in today's world where so much stuff is available online and most of the customers that are here today and Facebook has an engineering blog and pretty sure Yelp keeps an engineering blog and I'm sure Dropbox does too and others. So the accessibility that a smaller company that might not have an internal development team but still has access to understand how these technologies are being deployed, I just think that's a win for both sides. Well, we're at, I just got an invitation to remind you to go to the Facebook Developer Conference F8 Conference, April 30th. And you know, the theme, Zuckerberg is going to get out there, big, big to do, but basically their theme is build, grow, monetize, acre way, which is kind of their rhetoric. But that's ultimately the state of the market right now. So I want to get you to the final word and share the folks, you know, how are people really doing this? How are they building? How are they growing their infrastructure? Ultimately, they want to get to an environment and state where you got IT on premise in the cloud, rolling out applications in cloud under the hood with really, really integrated operating system like systems like feel. So how do people get there? Yeah, so I think the first thing is to recognize that, you know, there's no better time than the present and we've seen the model proved about certainly coming from the FusionIO perspective where flash can help people in terms of maintaining better consistency with their applications, better performance, but also, you know, shrinking the data center. If there's one area where I believe you get a common sense of purpose in technology is everybody wants a smaller data center. This is a challenge. This challenge is not going to go away. This size and scale of these problems are not going to shrink, they're going to increase. And so anything we can do to help customers be more efficient, both in top line performance and consistency, but also the bottom line and the dollar savings and shrinking that footprint and allowing them to scale more, but in smaller amount of infrastructure. That's happening every single day and there's dozens of case studies on the FusionIO website that can be viewed. You can actually now sort our case studies by application. So find the performance indicators that are relevant to your business. I think the eBay example of identifying for them what really mattered was URLs served per kilowatt or web pages served per kilowatt and they're now optimizing for that. They're saving money because they've found the right performance indicator for them. I think that, you know, every company has an opportunity to rethink the way that they're serving data, processing data, presenting data internally and there's lots of opportunities to increase the efficiency there both with flash solutions but also a number of other solutions that are here at the show. We're here at Gary Arnstein CMO of FusionIO. Congratulations on your announcement with MariaDB and Percona on the new flash aware non-volatile memory, obviously 4x improvements. You know, that's part of the industry cadence. You know, keep on reducing costs and increasing performance. Exactly. And obviously you work with SkySQL which we'll be talking with a friend in the second, Roger, a CUBE alumni shortly. Congratulations and, you know, we hope to see more accelerated action under the hood, the engine of innovations about data, data first and we're here at the CUBE. We're live at Percona, live in Santa Clara, California, Silicon Valley, right back after this short break.