 OK, thanks a lot. I'm really happy to be here. And my main objective today is going to be to introduce what we call the Oracle NoSQL database product. We thought, let's have a name that's actually descriptive. And you can tell from the name what the product is, as opposed to some of these other names out there. So I run the database development group at Oracle. Of course, we do the Oracle SQL relational database. We also do a number of other data management products, Berkeley DB, Oracle times 10. And today, we're going to talk about our latest product, which is the Oracle NoSQL database. And before I get into what we're doing with NoSQL and the NoSQL database product, I thought I'd give you a bigger picture of what we do in my group in an Oracle for data management. Because as you all know, NoSQL technologies and relational database technologies are very complementary. I'm sure most of you are using a whole wide range of these different technologies. And in fact, just out of curiosity, how many of you in your organizations use Oracle's relational database? Just to show you a good number of you. How about MySQL? Almost as many. OK. So what we are doing is a little different. We're not going to talk about why NoSQL is better than relational and why you shouldn't use relational anymore. We're going to talk about why you should use both. And moreover, why you want to set up products that actually are a vendor. They actually deliver multiple data management solutions that can be integrated together. They can be managed together. It can be supported together, as opposed to like, oh, no, throw everything out and only use my product kind of thing. So first, I'm going to talk a little bit about relational database stuff. Since I read a lot from what some of the NoSQL vendors are saying, and I feel like I have to sort of explain exactly what relational databases are good at. And then we'll talk about NoSQL and drill down into that. So let's go to the right side of this slide. What are relational databases used for? And by the way, the word relational database is sort of archaic now. Oracle and most of the other relational databases are now object relational databases. They're not just rows and columns of strings and numbers. They store very rich content. So the big, big market for relational databases is running the data management for a business. You're storing all the crown jewels, all the key operational data for running a business. And so you can go through all the different industries and you can explain what it's used for. In retail or e-commerce, the biggest, baddest e-commerce sites, Amazon.com, they've been using Oracle relational database since the beginning of the company to do their transaction processing to book their revenue. When it comes to money, you want an asset database. You don't want to lose track of your money. So we're very big, of course, in transaction processing for e-commerce. You go to other industries like telecommunications and banking. Again, this is all about the money and you don't want to lose any data asset transactions and SQL relational databases are critical there. The other big industry or big market segment that relational databases are very successful in is the BEI and data warehousing space. There, of course, customers want to issue SQL queries and they want to get a response back in a few seconds. It's not batch processing. They do batch processing also. But the big thing relational databases do for the data warehouse space is they can run on scale out clusters of big UNIX servers or commodity servers and deliver interactive response to very complex SQL queries. And it's all interactive. There are all these tools that on the fly generate dynamic SQL, send to the engine, get an answer back real fast. And the reason I wanted to mention those use cases is because those, unabashedly, if you look at the NoSQL products, that's not what the NoSQL products do. And so I'm always sort of puzzled about why they have to explain why they're better than Oracle or something, but whatever. That's what relational databases really do. And then they do lots of other things as well. If you look at the business applications that people use to run a business, they store all kinds of data in those databases. They store unstructured data. Relational databases have this data type called a lob and you can store anything you want in that lob. It can be XML, JSON, whatever you want, et cetera, et cetera. So big mission critical, highly available, highly secure systems, big massively parallel systems for data warehousing. The other thing I'll just mention is relational databases, the reason they're still vibrant after all these years is that the engines underneath the SQL layer are constantly being reinvented. As we go from one generation to the next. We're now moving into the cloud generation. We just completely re-architected the Oracle relational database to support multi-tenancy so that software as a service vendors can very easily take all their existing applications and have each one of their tenants run in what we call basically a pluggable database or a virtual database inside an Oracle database container and therefore give people the multi-tenancy or sharing of databases they want in the cloud. The other big thing going on in relational database now of course is in memory. Everybody's re-architecting their engines underneath the SQL layer to add in memory capabilities again to improve performance of course. So all the relational databases that are constantly being reinvented and this word new SQL applies to all the SQL databases, right? So let's go to NoSQL now. So actually MySQL of course is another relational database we support, it's not quite as mission critical or scalable or available but it's quite a popular database for departmental and web applications and they've actually just recently had a new release where they have added elements of what people like about NoSQL to MySQL as well. So let's go to NoSQL now. So let's just focus on that for the rest of the presentation and of course Oracle acquired a company called Sleepycat about 10 years ago. Sleepycat built a database called BerkeleyDB. BerkeleyDB is probably the most popular key value stored database out there in the market. It's open source, very successful, especially in embedded systems. It's inside a lot of systems you don't know is there but it's there and then more recently we've come out with our NoSQL database and the key thing we did there is we said, hey BerkeleyDB is a good key value store for a single node but if you wanna scale out, let's use that as the building block and that's what we did and Oracle NoSQL database is essentially a collection of a distributed key value store across a large number of BerkeleyDB nodes. And so let's go a little bit more into what we're doing with NoSQL. Of course we're doing what everybody wants in that space. We've got schema support, key value store, relaxed consistency, you can have acid transactions or not, et cetera. I'll go into that a little bit more on the next slide. So just at a high level we thought, this is not a real technical presentation. We're gonna have some drill down presentations this afternoon and tomorrow to go into a lot more detail. But I thought I'd give you just the flavor of what we're doing here. It's a classical distributed key value store. You've got what we call the smart topology driver which is a smart client that obviously looks at the key and hashes the key and sends you to the appropriate shard and then the appropriate replica within the shard for that key, right? And it's a smart layer. It does load balancing. It understands data center. It understands the distribution of nodes across data centers to make sure if a data center goes down, you've got a replica and a surviving node. Load balancing, what we call multi-master white partitioning that we'll talk about in some of the later sessions. It's elastic, you can very easily add shards and nodes and storage all online with very little impact on the performance of your system and we'll actually show some numbers around that. It does auto-rebalancing as you add shards and nodes. We under the covers are moving data around for you without you knowing it all online. It's a very cool technology and we have both base and asset. And I mean, a lot of people say, oh, asset is bad or I don't quite understand this but when you have a failure, you're doing an update to some data and the storage fails, you don't want your database to be corrupted. You wanna be predicted by something like a logging infrastructure and that's what we use in the Berkeley DB core of the product to make sure your data is always correct and not corrupted or inconsistent. But we do, of course, support eventual consistency in that kind of model but we know you don't want your data to be corrupted so we also have rowback in the event of failures. Okay, let's go through a couple of use cases and one of the themes that's a couple of these use cases is again the complementary nature of relational databases and no SQL. So go into a banking situation, for example. So you've got your credit card and you wanna buy something. So the first thing they wanna do is they wanna make sure this is not fraud. You haven't stolen the credit card, right? So what they're gonna do is they're gonna use something like a key value store and we're working with some vendors, of some customers right now in this exact use case and they wanna pull up your credit history, your transaction history and they wanna run some scoring algorithm against your history to make sure this looks like a good transaction. So this is a very classical key value store use case. You're pulling back content, it's read only, it's very scalable, you can take advantage of the scalability of key value stores. You know, this algorithm was probably created by a relational database and a big data warehouse somewhere where they're running SAS or R and computed the scoring algorithm for deciding if something was fraudulent. So there's a relational database back there probably creating that algorithm for scoring. But the no SQL engine will do the scoring and then when they decide, yeah, this is a good transaction, then they hit the relational database, they do the transaction and let you buy the product, right? So that's classical banking use case. Manufacturing, again, there's often a lot of sensor data and manufacturing situations. Another good use case for key value stores, you can ingest that sensor data really fast. You don't really need transactions, you lose some of the sensor data, it doesn't really matter, right? Just throw it on the floor. Retail, another classical example of integration of key value stores and relational databases, you know, back to internet e-commerce, Amazon.com example, for years and years Amazon has been using Berkeley DB or key value stores for products, for product catalogs, you know, when you go to the Amazon store and you're looking at the products they've got, that's usually hitting a key value store. The shopping cart might be a key value store, right? But then when you finally go and say, hey, I'm gonna buy something, that's when they hit the Oracle database, they do the transaction, they do the purchase. So again, very complimentary technologies, key value stores and relational databases. Okay. Okay, so what we're gonna do in the next few slides is go into three separate areas where we think we have a really strong offering in our Oracle NoSQL database, primarily around availability. You know, we have worked for many years in making these mission critical OTP systems highly available and we're taking some of that expertise, we're putting it, applying it to our NoSQL technology and we think we're doing some things there that I think we think are gonna be very differentiated versus a lot of the products out there today. So we'll go through three areas and let's just go into the slides. Okay, so the scenario is it's October and in a few months it's gonna be Black Friday, Christmas is coming and you know you don't have enough capacity in your key value store and you wanna start adding that extra capacity so that when Black Friday comes, you're gonna be ready to go, okay? So what we're doing here is we have an experiment. We had 144 of these of the nodes of the NoSQL database running and we wanna grow by 50% capacity, we wanna go to 216 nodes and we wanna do it online, okay? And so what this experiment shows if you look at the Y axis on the left, we're showing throughput. So these bars are the throughput of the running system and we're showing on the X axis sort of how far through the growth from 144 to 216 we are, 20% through, 40% through, et cetera. And then on the other Y axis, we're showing response time for reads and writes and that's in the line graph in orange and red up there. Okay, we start out, we're running about 50,000 operations per second. We start adding the nodes and we go through about 20% of adding the nodes and this is where it's adding shards, adding replicas, all that kind of thing. We're rebalancing data across the shards online. And you can see there's a little bit of a performance hit. We went down from like 55,000 to 52,000 briefly for the first 20%, but you can see it's bringing up the bar chart by 40%, it's back up to the same throughput. And then as you get towards 216, you start seeing the benefits of the extra nodes, you see extra throughput, you see response time, latencies going down, okay? So you're getting that extra node, extra capacity, okay? But so far you haven't changed the number of clients. So the workload has stayed flat, okay? So now let's say you bring on the extra clients, you bring on extra workload and that's what we did in this experiment. We now went from 360 clients to 540, so we grew that by 50%. And what we just wanted to show you here is that we have linear scalability. We started out at 144 nodes, we now have 216 and we scaled with the extra client requests from 52,000 up to almost 80,000. So pretty much linear scalability. And again, this is all being done online, no downtime while your production system is up and running. The third area we wanted to show you is rolling upgrades. Now, what's this about? Well, unfortunately, all products have bugs. You have to apply patches for those bugs. There are also, of course, new versions of the products that give you new features. And the key thing is when you have these big distributed key value stores with hundreds of nodes, you don't wanna, you want your system to be up number one all the time. And you want the process of applying these patches to be really fast. And so what we show here is at different sizes of clusters from 72, 144, 216, how long does it take to apply the patches in a rolling fashion node by node without taking the system down? And what we're showing here is with 72 nodes, we can do it in about five minutes, 144, about 10, 216, about 15. So again, you can see we're scaling linearly. Again, this is all being done as the system is up and running, no downtime. So we're very excited about all these three capabilities. And we think this is the kind of technology customers really wanna look for in the NoSQL space. It's not just about being able to develop really fast. It's also being about once you go into production and you're running this thing seriously to run your business, you want a product that stays up all the time and is not, you know, it's like, okay. Okay, so what we just wanted to say here is, you know, you're using a lot of different data management technologies in your organizations. We saw from your show of hand, you've got lots of Oracle databases. You've got lots of MySQL database. And now you're looking at NoSQL databases as well. You want a vendor who can help you not only support all those products, but also help you integrate them together because once you have data all over the place, the next thing your business wants is they wanna understand what's going on with that data. They wanna integrate it together. They wanna analyze that data. And in order to do that, you need a combination of data management technologies. NoSQL is good for what it does. Relational is good for what it does. And you wanna be able to merge that data together. We also mentioned Hadoop here. This is, I'm not really gonna talk about Hadoop today, but let it be said as part of our whole big data space, data warehousing is moving into the big data era. And data warehousing and using a relational database for massively parallel SQL, plus the Hadoop platform for doing interesting ingestion of all kinds of data that people call in the big data space and integrating that in your data warehouse is a really important task. And we're also, of course, integrating data across Hadoop and our relational database as well. Okay. Okay, so just sort of to summarize again, we believe the right way for what customers want is they wanna use the right data management tool for the right job. That means they're gonna be running relational databases for what they're good for. They'll be running key value stores for what they're good for. They're gonna be integrating data across those different databases. They want a single management framework and Oracle has an enterprise manager product that integrates all of our products, all the data management products, also all of our middleware and application products as well. Oracle support can be your single place to call. If you've got a problem, you don't know which layer of the stack it's in. You can call us. We will take care of the whole stack, whether it's in Berkeley DB, Oracle NoSQL database, MySQL, or the Oracle database, or any of our middleware or applications products as well. Okay, so let's get to what we're announcing here today. So we held back some things that we've been working on for announcing today, because we thought this was a great audience to talk about some of the new things we're doing. The first thing we're doing is sort of a business model thing. So obviously Oracle databases and all of the Oracle products for years have been sold via what we call perpetual licenses. You pay us money for the license up front, okay? You get the use of that software for the lifetime of your business, basically. And then you pay us an annual fee for support, right? In the open source space, people have been using a slightly different business model which says, okay, you don't pay us up front, but you do pay us annual support fees for getting support for our products. Or we're just open source, you can just download the software from the web and play with it for free. So what we're announcing today is we are going to add to our existing perpetual licensing business model for Oracle NoSQL database, a more open source friendly kind of model where we now have a version of Oracle NoSQL database called the Community Edition. That edition of the product is available for download by anybody on Oracle's website. You go to OTN, which is what we call our technology network. You go down, download the software, you go play with it, you don't have to pay us anything. And then if you decide, hey, this is a pretty cool product, I wanna get support, what we're announcing here today is now you can get support and you can get it in a very simple fashion. You can just go to the Oracle store on the web, shop.oracle.com, you can give us your credit card number. And for $2,000 per server per year, which is actually a very good deal, you get full support, the full enterprise class support that Oracle provides for all of its other products are now available for the Community Edition of our NoSQL database product. And I know a lot of you maybe from startups, startups of course don't wanna spend any money at all, this is great for them. They can just start using the software and then when they finally become real, go live, they can get support and get the full benefits of Oracle's support organization. Okay, so that's the first announcement. Number two, we wanna talk about appliances. So over the last few years, Oracle, since we bought Sun, is now not only a software vendor, but we're a hardware vendor too. And what we've been doing is something we call engineered systems, other people call appliances or converge solutions. What we do is we take our software and we take the hardware, we integrate it together. We're basically like a systems integrator in my group. And we deliver appliances to customers so that customers don't have to be in the business of engineering, complex, clustered computer systems anymore. The first one we delivered for my group is called Exadata. Exadata is sort of a play on the Teradata name. And Exadata is this database machine for running massively parallel, clustered, scale out database technologies, database systems for big data warehouses. We have customers running over multiple petabytes of information on the Exadata systems for data warehouses. They're also used for big transaction processing systems as well. After that we added something called the Oracle Database Appliance, which is again an engineered system for the Oracle Database for more low end use cases. And what we're announcing today is our big data appliance for Oracle NoSQL Database is our latest engineered system. So if you wanna go and start playing with Oracle's NoSQL technology, you can buy our big data appliance. You can start with a six node cluster of two socket commodity Intel servers. And that's what we call the starter rack. And we will have a whole set of scripts to pre-install your software for you on that machine so that you can be up and running in a matter of hours. You order the hardware, you order the software, you load it up and you're ready to go. So it's something we found our customers really like, this whole notion that they don't have to design complex clustered computer systems anymore. Okay. Of course, all of the technologies I mentioned earlier work in this environment, Enterprise Manager works, we give you full support. The big data appliance also is an engineered system for Hadoop, so you can also run Hadoop on the same hardware as you run Oracle NoSQL. You can have some of the nodes on the big data appliance running Oracle NoSQL Database. Some of the nodes can be running Hadoop if you wanna integrate the two together. We support that as well. And just the last note, if you wanna go to this link, this hardware is very competitive in price and if you do the math, we think you'll find in a lot of cases this is actually cheaper than going out and buying hardware yourself and integrating it together. Okay. And then the last thing I wanted to mention is that in order to get people playing around with our product, we welcome you to go to this website, noSQLcontest.com and play around with the product, come up with some interesting app, send it into us and we are gonna give a prize, an iPad is a prize for whoever sends in the most creative use of Oracle NoSQL Database. The contest ends October 18th, so download the software today and start playing. And with that, I'd like to welcome you all to, go to our website, get more information. At the conference this week, this afternoon I think we have a panel session from some of our developers who are gonna be here to talk about Oracle NoSQL Database. I think we have another session tomorrow as well, so where you can learn about the technology in a lot more depth. And with that, I think we have a couple minutes, I'll be happy to answer a couple questions if we've got any questions from the audience. Sure. $1 per server offer, is there a limitation on the number of cores? So the question is, I said the price for support is $2,000 per server. And don't recharge per core. So Oracle for our other products, we do license by the core or processor, but for the NoSQL Database the metric is per server and that is because that is what is the sort of standard metric people are using in this space. So we're just following the same metric, all the other vendors in the NoSQL space are using. So we have no per core charges, it's purely per server. Raise your hand if you have a question. Any other questions? I had one for you actually. So Oracle also has some graph capabilities in your databases. Is the Oracle NoSQL Database, in terms of the typical NoSQL categories, document stores, key values, are you covering all of those bases or do you put it into one of those slots in particular? So the question is about graph technology, so... Well, not specifically back. I'm saying you've got graphs already. Now you're introducing the NoSQL Database. Which of the NoSQL architectures is covered by? Well, the Oracle NoSQL Database is a classical key value store. It's built on Berkeley DB, so the keys and values are just strings of bytes, right? And you can store JSON documents in the value if you want to. You can store... We actually did an experiment recently where we show you how to store RDF graphs in our key value store as well. The Oracle Database also stores graphs, and we support RDF and Sparkle and some of the standards in the graph space as well, and we're doing parallel SQL extensions to do graph traversal too. So again, you can use both technologies for different use cases. The NoSQL product, of course, is a great... is a store. It's not trying to do massively parallel graph processing, but on top of it, you can obviously run those kind of algorithms there as well. Two more minutes. Okay. The NoSQL supports OLAP or cube-like structures where analytics can be done on the NoSQL directly or the data has to be moved to OLAP to do analytics work. Yeah, so the question is about analytics. Now, in my book, analytics is one of those spaces. The BI data warehousing space, where SQL is actually a very cool language for building very relatively simple queries that are the equivalent of hundreds of lines of code. And so if you're doing analytics, generally you would move the data from a key value store into a relational database where you can use all the world's BI tools to analyze that information. So I think today, at least, our sweet spot is not analytics necessarily, although you can obviously write code, write Java code or whatever to do analytics against the data in the key value store. But if you wanna use the tools, there's a rich set of BI tools, of course, that run against relational databases that might be a more productive way of doing analytics. Okay. Thank you, Andy. Okay. Welcome to, thanks. All right. Thank you very much. Thank you.