Autonomous Data Management: Andrew Mendelsohn at Oracle OpenWorld 2019

the Oracle autonomous database the world's first self-driving database sir relax everything's automated human error is eliminated a revolutionary new technology called machine learning enables the database to protect and repair itself without ever having to stop the database automatically responds to cyberattacks and prevents data theft in a world striving to build self-driving cars Oracle has built a self-driving database fully automated ever-vigilant your data is safe [Music] in the auricle autonomous database cloud [Music] please welcome to the stage executive vice president database server technologies Oracle Andy Mendelsohn okay good morning everybody it looks like you've got a full house here this is terrific and what we're gonna do is we're first gonna have Pat Sullivan come out here Pat is runs Accenture Oracle business in North America on out here Pat thanks Andy and Pat is gonna talk to you for a few minutes and then I'll come on next all right all right thanks Andy all right well to kick this off I want to tell you a story so 22 years ago I'm sitting in my dad's office and put this in perspective my dad's uh he was a University basketball coach and there was this big monitor thing on his desk and I said dad you got a computer on your desk my dad looked at me and he just had this huge grin on his face he was so happy and he goes Pat I haven't even used it and it's been here for seven months hey all right well dad look you know between my four sisters and I this is a way in which you can get in touch with us in the way that we're communicating this through this thing called email and so he got a little excited and he goes all right well help me set it up so I put in all my sister's email addresses I put mine in I put the keyboard in front of my dad and I say okay let's get ready you can you can send your first email so my dad's sister and he looks down looks back up at me he does it again he goes Pat why aren't the keys in alphabetical order so now that that same man my father he's got a smartphone a tablet two laptops and he just finished publishing his second book so if my father the human can make amends and embrace technology with the machine than so can todays enterprises in this post digital area that I'm going to talk about so our destination today is autonomous and over the next 20 minutes I'm going to take you on this journey where we explore what's happening with data the trends that are going to create more data and really put this data to work to drive new revenue streams competition and new businesses and so right now we're in this middle of an explosion of data and you know there's always been too common data sources there's been consumer data and then there's been enterprise data and these used to be pretty obvious until kind of what's happening today and each of these are growing at unprecedented rates today and so there's a statistic out there that says we're generating over 2.5 quintillion bytes of data every day have you ever wondered what 2.5 controling can quintillion looks like well check this out it's a 2 a 5 and a whole bunch of zeros and to put this in perspective let me give you a couple of points on just what big numbers look like so a kilobyte okay a kilobyte is a thousand bytes that translates roughly to about 17 minutes ago and I was backstage having a conversation with Andy a megabyte that's a million bytes and I was 12 days ago and I was on my computer booking a flight to be here with all of you now a gigabyte which is a billion bytes I was 32 years ago and I wasn't here I was in school I was in sixth grade now if you go to a terabyte or a trillion bytes and put that against time that was thirty one thousand seven hundred and ten years ago and I think that's probably when my dad was working to invent the keyboard so to get to a quintillion you need 2.5 million terabytes or millions trillions and that's a lot of data and that's data that we are creating every single day so the question you might be asking yourself is where does all this data so let's take a look at that so I call this the the data minute and since I started talking about six minutes ago you know let's look into what happened in one of those minutes so if we look at this and explode this out nearly 60 million text messages were sent in a minute between messaging apps and whatsapp but a hundred and eighty-eight millions emails were set what's interesting about that is a hundred and four million of those emails are estimated to be spam more gifts were served up than searches on Google and more people swiped left and right on tinder than people who logged into Facebook but most importantly over a million dollars or just about a million dollars was actually spent online in a minute so when you think about the the data that's coming from that it's pretty monumental and most people would consider what's on this slide to be consumer data but not companies that are taking this data putting it to work and personalizing the data to create offers individualized for us they would consider this the golden data and that's part of this new world that we live in both as consumers and as enterprises so the best way to discuss this is actually through this report that our company Accenture does every year called the Accenture technology vision so we've been producing this vision for over a decade it comes out every year and in each tech vision we look at different trends that are shaping out in the enterprise some of these trends are a few years out while others are happening right now and when we look at these we we try to understand what innovations are really going to drive these trends to to work out into the market and the tech vision for Oracle is something that we published just about five months ago and so this year's tech vision for Oracle is centered around the emergence of this post digital era and how we're moving beyond the adoption of these digital tools to a new generation of technologies and we'll talk about autonomous and how these technologies and innovations are applying to the business where businesses can differentiate themselves within the marketplace so very simply digital technology is no longer this competitive advantage that we used to talk about and help companies achieve it's the price of admission and this new set of rules for businesses in this post digital world it requires the use of powerful new technologies to innovate the business models personalized experiences for customers and customized products services and even people's surroundings for example with extended reality all on-demand but with these new rules come new and deeper responsibilities so in other words post digital is about these three tenants individualization instant on-demand and momentary markets these are three critical elements for businesses in a post digital era and they're all backed by guess what data and so we surveyed over 6,500 executives from companies over with revenue over 6 billion dollars and here's some interesting findings so 85% of these businesses they believe that customer demands are moving their organizations towards customized and on-demand delivery models the same number say that integration of customization and real-time delivery is the next big wave of competition but what they all said is that it's going to take them three and a half years to get there and in order to deliver this on demand these custom products any services in real time so to differentiate in this the merging post digital market companies have to understand the combination of being able to know people at the holistic level where they can instantaneously understand demand but then be able to meet that demand instantaneously again it's all about data and these will enable companies to deliver his needs change at a moment's notice so example of this is actually how Adidas and Siemens are working together adidas is on this initiative where they want to customize every single shoe so you can go to a DS com you can customize colors different logos where things are but they don't have the warehouse in order to do that and that's where Siemens comes in so the two companies are working together to create these things called speed factories to be able to customize and do this on demand and get shoes out to their consumers all right so we know that data is exploding in a tech vision characteristics certainly showcase a data's being used as this weapon for disruption in a creation of new revenue streams and in order to access manage store secure analyze and integrate all of this data it requires a deeper look and this is where we test the value of humans working with machines so every day at Accenture we've got a group of close to 20,000 Oracle data professionals across the globe and they're helping our clients process over 60 billion transactions over three exabytes of storage on oracle databases and so as a result of the work that we do here Andy and team reached out to us two years ago to beta-test the autonomous data warehouse and we really liked what we found and so in our testing we're able to insert 500 million rows in three minutes and create data Mart's in mere minutes without using core Oracle DBA skills but that was two years ago so given the explosion of data and a continued maturity of Oracle's offerings our Oracle engineers at Accenture decided to put autonomous to the test and so our findings are really simple and I'm going to go through this for the next few minutes but Oracle's autonomous database data warehouse our solutions that thrive in this post digital world we're launching this new report at the end of this presentation that you can download and read which goes through all the results that I'm about to go through all right so there's three key pillars that we're going to go into the first one is speed does it perform the second one is money is a cost effective and then the third one we measured is value will it deliver business results so should we jump in all right let's do it okay so speed testing performance honestly can be as much of an art as it is and our focus was not to break records when we did this testing but to simulate what we typically see in a work that we do for traditionally fortune 500 companies and so there's a few key details that went into our methods before I go into what you see on her so what we did on the first two configurations the leading cloud provider and OCI is we installed Oracle database 19 C in a concept that we had here is that most of our clients have their own licenses so we took a bring your own license approach to this now what we could not do is install Oracle database on autonomous and by the way if we could it wouldn't really be that autonomous but we accepted the installation of what Oracle chose for 19 C on autonomous despite what our Oracle aces what would really like but keep in mind that when you do the installations that takes time with autonomous it just happens and it happens in minutes so we also use swing bench swing bench is pretty much what the entire industry uses to generate loads and performance testing for Oracle database workloads and then we have an OLTP data set that we tested with that generally we use in a lot of our clients so pretty active lots of transaction production type grade data set and then we ran these tests on the three different clouds over 50 times for an hour and we did this at all times during the day so we could watch any variations based on cloud usage or network or that concept of the noisy neighbor effect and then we took the averages of all of these tests and that's what we published in our research paper now we tried to make each of the clouds that you see behind me look alike but not all clouds are alike in a shapes that we can buy so each configuration had the same number of CPUs and remember a two CPUs on a non Oracle cloud is equivalent to one Oracle CPU or OCP you okay now memory was fairly similar however that's why rosy the robot from Jetsons is up here because on autonomous that's automatically allocated for you and it's just done again it's autonomous and so we selected solid state disk on the third party cloud we also did that on OC I and you know this is where the the storage or the disk size is where we're limited so on OC I the smallest shape you can get for disk is six point four terabytes on the nvme and then the last thing is we've implemented thousands of exudate as for our clients and so we knew with autonomous because it's Exadata in the background that this was gonna kind of be like a an Indy car in a stroller derby and it was gonna be a little bit of an unfair test if you will so what our engineers decided to do was add another shape and so we did this with a 2o CPU shape just to kind of test our hunch and to see what was going to happen so with our testing environment set we ran a test over 50 times for each hour and here's what we found the non Oracle cloud completed 65 queries per hour OCI 19 times more queries with the same exact workload and then the autonomous data warehouse just screamed by running a hundred and eighty four times more queries with the smaller configuration of autonomous running at 37 times so essentially here we have proved that the autonomous data warehouse has phenomenal performance but the question we ask next is what is that going to cost us so let's look at the money slide so there's a lot of information on the slide and the first thing that your eyes are going to go to is that bottom row and looking at the total cost but let me take you through a few things the cloud prices here are published prices based on the configurations and the terms that we selected we assumed that bring your own license for the Oracle database but there's still support costs that go with that from an annual basis and so the support costs on OC I and the other cloud we base that on 19c with advanced security is the only option an autonomous when you bring your own license the infrastructure and the support costs are built into that service offering but here's there's some some key points regardless of how many queries these clouds can run the pricing represents the total costs for running Oracle database and mission-critical OLTP data sets an autonomous data warehouse includes all database options that Oracle have including high availability or RAC so if you wanted to add compression or partitioning or RAC to the other two configurations your price gap starts to widen significantly favoring of course autonomous so time like I stated before is not factored in here that time to do the installations on the other two clouds or to create ansible scripts if you will but you've got to update those every time new patches and releases from Oracle come out so what you see here is autonomous is clearly price competitive with the total cost being 13% less on that larger configuration and then on a smaller configuration two and a half times less than the other cloud so we've proven that Oracle's autonomous data warehouse is really performant and a total costs is lower than other clouds where you can run Oracle database workloads so now I want to look at the value and I want to kind of put these two together and let's see what we got so here's the value so how much do you want to pay for every transaction that you run is really the question so if you expand out the total cost and the queries by our we wanted to understand what it would cost to get the work done as we all want to spend less time to do more and again remember what we talked about data and the growth of data to be able to query that quicker could give you a business advantage and so if you just think about data I've been up here now 15 minutes or so and in that time we've added five hundred and twenty thousand new terabytes of data just in that time since I've been talking so being able to run more queries could be the difference between your company being disrupted or your company disrupting a competition so we know the number of queries we know the speed we also know the annual cost per configuration the money so we explored the value of running these queries by hour and by query as you can see up here so the total cost to run OCI with the same configuration is roughly half of the other cloud or 97% cheaper when you look at it is total cost by query and if we were to look at this from a different perspective the autonomous data warehouse did not let us down so the cost per query on a large configuration is 200 times cheaper on a smaller autonomous configuration this is really interesting it's only 1% of the cost to run that compared to the other cloud when you look at it per query so another way to look at this which could make a lot of sense for all of you certainly for me is literally when you put your money on the table and you look at it so let's take at it from this perspective so it makes a lot of sense if you will when you look at it from this area and for each query on the leading cloud provider you're gonna you're gonna pay 20 cents per query ok on OCI it's half a penny on the smaller configuration of autonomous it's a quarter of a penny and then when you look at it then that larger configuration getting work done it's a tenth of a penny so it's pretty phenomenal what what Andy and team Herot Oracle is done so if you're an enterprise and you're leveraging data to help your business grow you really should be leveraging innovation that's going to yield the largest results as we show and these are tests using data sets as well as what we do with our clients in that production manner all right now there's one more thing here I want to talk to you about and this is that patching for security so if you went out in googled what it would cost for an average data breach in 2019 you're going to see this thing three point nine two million dollars on average so one data breach three point nine million dollars right that's just the average just think of what it would cost for your enterprise and obviously bigger enterprises potentially a bigger cost so we believe it Accenture that securing your data is not only critical it should never be a choice and what we like about autonomous is that it defaults to security right with the solution and so anytime Oracle sees a potential threat that comes in with their Cloud Security Operations Group they immediately start to apply that threat into a patch and the first thing they do is they dynamically update autonomous but then as all of us know if you installed Oracle 19 C even on infrastructure as a service you're still liable to have to patch that so enterprises can take days weeks and we've seen months to apply these patches and so if you think about that with the sensitivities around not having that patch and what it could cost you you really can't put a price tag on doing that so this further adds to the total cost which is hard to realize and put up here but you know we do believe that being able to remediate that potential loss of business as soon as it happens you know is important so any loss of customers data generally results in a business loss as well okay so I want to continue this and build on this so these were the three characteristics of our tech vision and you know with the three characteristics in a stellar performance results that we saw from the research that our team did with Oracle's autonomous offerings it's really important to understand how this is actually being applied and so trust in responsibilities is a key thing whenever you're dealing with vast amounts of data so data served up by machines right but we as humans need to be making sure that we're using it responsibly so I'm gonna give you a few ways how Accenture is applying these autonomous technologies to solutions that we implement our clients across three major areas all right so the first one is what we look at with industry solutions we have a solution called the Accenture life sciences cloud and this is a solution where we aggregate clinical trial data and most of the large pharmaceuticals use this solution today and so basically the data is flowing from all sorts of different source systems into one platform and what this allows the pharmaceuticals to do is use data as a way to create new applications to the different governing bodies such as the FDA in the United States what the FDA has done is they've now allowed the use of historical data to be used in these submissions to essentially improve the times in which drugs are approved so the ability to analyze this data through complex algorithms and produce the scientific scientific data is why we're why we're leveraging autonomous in this solution you know the data certainly needs to be trusted but we need to be responsible with it and when lives are in a line that's pretty important the second example is back-office transformations we're doing a lot of these at Accenture today with our clients and in transformations where clients are moving or back office applications to SAS Accenture we're offloading this historic data which at times could be anywhere from 15 to 30 years of historic data and we put this on autonomous and we do that because it works and the management of this platform is essentially hands-off we can focus on transforming our clients to what they need and so this is rich enterprise data however that's never really been leveraged in a way in which we can personalize it with other data to create these new offerings for other businesses and clients that they have so again this data also needs to be trusted but we we have to be responsible with it now the third one is what we call the migration of the old we're starting to see a trend right now with our clients where they're offloading their legacy data warehouses onto the autonomous data warehouse so Hearst Communications is one of our clients they're a mass media business headquartered in New York and you know the thing I like about hearse is they're not waiting to be disrupted they're moving their data to a platform that can query it faster than anything else can to create these momentary markets and believe me that matters at Hearst based on a business that they have and again this data it needs to be trusted and we have to be responsible with it all right so let me wrap this up for you I can clearly say that Oracle runs best on Oracle cloud Andy's team has done just a remarkable job of bringing great innovation to the market and bringing solutions that can take on this massive growth of data and so you know the thing I'll leave you with is this Andy's team they make excellent products and it's Accenture job it's my job to make it work for our clients so I've got one question for you are you autonomous give me a call our team will help you get there all right [Applause] all right but thanks so much you can actually download the study at the link care and also our tech vision and right now I'd like to bring Andy back out to take you through all the great innovations that his team has been working on in data management and the next generation of autonomous ok thanks Beth thanks Andy ok so let's get going and what I want to start off with is just a little bit of a discussion about what it is that we in Oracle database development are focused on these days and you know everybody thinks Oracle's just talking about cloud and that's all we do is cloud cloud cloud but it actually is not exactly accurate so the the big thing we are focused on number one is today I think it's very well recognized that we have you know the best database software technology out there and our focus is that 10 years from now we want that still to be the case we want not just be the best we want to be you know 10 years ahead of everybody else so we have a huge focus in just building core software for database software that runs everywhere runs on premises it runs in our cloud and runs on other clouds and you know that's number one priority we want to be in the lead for pure database technology for the long term number two of course we're very focused on engineered systems we we started working on x8 about 15 years ago it's pretty hard to believe but that technology is really important on Prem you know you saw from the last presentation some examples of the performance of exudative based infrastructure that we use on our cloud and of course you can use the same infrastructure on-premises as well and that technology one of the big things we're doing there is and there's gonna be announcement later today as we are making that technology more cloud need you know today in our cloud there are these boxes exudative boxes but a year from now EXA day is going to be more like a virtual why's technology where you can push a button and on our cloud you'll get a virtual Exadata that's composed on the fly out of the component parts the storage and the networking and the compute servers that you need so we're doing a lot of work there and then finally of course autonomous database you all know about that you just heard a little bit about autonomous database in the last presentation and I'll talk about what we're doing there okay so let's just focus on the software innovation part first first a little review I think most of you understand we've gone to this new release model where we every year we ship a new version of the database and the name of the release is basically the last two digits of that year so 12c r2 is the last of the old-style naming that we shipped in 2017 and then 18 C in 19 C are effectively what would have been called 12c r2 patch set 1 and 12 cor to fetch that to with 19 C being the terminal long-term support release of 12c r2 so that's sort of where we are okay in the release process and I just wanted to review really quickly what we did over those last three releases because as customers of many view our work our using Oracle database releases before 12c r2 as you upgrade of course the 19 C and that's where all of you should be going now not any of these earlier releases you get the benefits of all the innovation and all these earlier releases 12c are two in particular had a lot of innovation in it the big new feature we did there is something called sharding and sharding is not for most of you it's it's mostly designed for a really extreme distributed transaction processing systems you know the original sharded systems were like amazon.com manually shorted Oracle databases the run e-commerce all over the world and that was sort of the early days of shorting now we've we've automated sharding so that you know you don't have to understand that they're you know a different database in each of the different locations in the world we take care of that for you there's a really nice use case for charting also if you are a global business and you have customers in different countries and the world and their regulations that say the data about the customers has to stay in a data center in that country I'm charting is really good for that too you can create a shard for each of these countries that need to have local data and you can you know follow the regulations around that we've done a lot of work around multi-tenant that was new and 12c r1 multi-tenant has become the foundation of everything we do on the public cloud autonomous database is using multi-tenant our fusion SAS applications all run on top of Oracle multi-tenant databases net Suites moving their Salesforce moving there and then on-premises customers are building private clouds I'm using multi tenant as well so it's a really valuable technology that's very flexible it fits really well with what people want to do in both private clouds and public clouds and then you know we're working on hansung all the technologies in the in-memory technology is continuously getting enhanced and that'll be going on for years online database encryption is really important you have a lot of data that's not encrypted everything needs to be encrypted nowadays and it's now very easy to do that because now while your production system is running you can encrypt your data in the background online without any any downtime so that's that's really nice feature in 12c r2 18c as I said is you can think of it as the first patch set to 12c r2 if the big focus was just on you know quality fix lots of bugs fix security bugs we did some some small number of features as well on that release you'll see things like you know Microsoft Active Directory Integration before you couldn't integrate the database directly with your Active Directory server you had to go through some intermediary that's no longer needed starting with a teensy and we did some other think private temp tables really nice for developers if they want to have a temp table just temporarily during the life of say a stored procedure or something like that that's a nice feature and then 19th C which is our long-term support release we actually did a reasonable amount of innovation in this release some of it was for cloud and some of it is just for typical use cases the customers have the first one was for IOT so with IOT data a lot of it's like sensor data data from Co car telemetry etc this data is not really that valuable so if you use lose some data points here in there nobody really cares it's not like a banking transaction or something like that so in nineteen C when you create a table you can now tell us it's one of these kind of tables that's collecting you know less valuable data and if you tell us that we will when you say commit well get the data disk eventually but the focus is really on performance so we'll buffer all your your loads up your insert statements up in memory and we'll eventually get a disk and we pretty much will do but if this if the system will crash we might lose a little bit of the sensor data it doesn't matter because nobody cares so the reason we do that is because in exchange for that possibly losing some data you can get to three times faster low performance without rewriting your code just standard sequel and Stewart insert statements against this kind of table will just run much faster active Data Guard updates this is a really important feature a lot of you are using active Data Guard you know it's for read-only sort of reporting workloads so you your standby database can now be valuable it's not just standing there waiting for the disaster to happen right so unfortunately a lot of these read-only workloads sometimes really aren't read-only they do a few updates and so if you have one of these read mostly workloads which is say you know 95 percent read 5% updates you can now run them against the active Data Guard standby and it's all transactional it's all transparent to your applications it makes active Data Guard much much more useful so that's a really nice feature automatic indexing I'll talk about later that's part of our autonomous database self tuning capabilities sequel query of object stores this is another feature that we built originally for autonomous data warehouse so that you could easily load your data warehouse by taking files putting them into the object store and then using standard sequel statements to insert data into the data warehouse by selecting it from the object store files it turns out customers are now and I'm sure many of you are looking at using deploying object stores on-premises object stores have become in a lot of ways a replacement for Hadoop HDFS it's a way of storing lots of data at really low cost and so we decide to make this ability to query the object store available on Prem as well and then finally Jason 12c r1 we added JSON support it's a course as you know very popular with developers we're continuing to Hance that and and we'll talk about what we're doing in the next release 20c around Jason a little later ok so why 19c as I mentioned this is the long-term support release everybody's been waiting for so if you're on 11g our 211 204 whatever and you've been waiting for the next big long-term support release this is it everybody should be certifying there there is V applications moving their their databases to 19 C it will be supported all the way out to 2026 if with extended support it's sort of a no-brainer this is at this point in time don't move to any of the erbium don't move to 12 e are - don't move to 18 C now is the time to go to 19 ok ok so in this presentation I have a few of these myths statements because when you go out there most of you understand Oracle really well but the the more junior members of your organizations the developers often you know they've grown up in this this world where there's a lot of Mythology going on around relational databases and you know they've been using these less powerful relational databases you know they're using Postgres or my sequel and all the limitations of those databases are thought to be limitations of relational databases but of course they're not they're just limitations of those products so one of the things I want to do today a little bit is debunk some of these myths so you're armed and dangerous when these these developers come and tell you how Oracle can't do this and Oracle can't do that you know well it's so the first one I want to mention is this one this is the one you hear from a lot of the Hadoop community the big data guys there's like okay we want to do all this these analytics against our data and of course relational databases can analyze unstructured data they can't do machine learning models etc etc yeah you know they can't scale and so this is the first one let's talk about this one so of course relational databases were purely relational 30 years ago you know when we just supported numbers and column you know columns with numbers and dates and strings but over the years relational databases have become multi model databases they Sui support relational models we support Jason as a data type we support you know XML spatial data graph data of course key value data R is a very simple subset of relational data so the developers can get what they want which is they want their popular open source tools to write code in all the languages they own in Java JavaScript Python etc we give them all that and we give them powerful open standard sequel to access all these different data types you know the popular cloud guys out there they give you all these basically they base encourage you to take your data and fragmented it across nine different cloud services one for each of these different kinds of data types because they don't have powerful database technologies that can really do multi model really well so we give you powerful query across these data types so in the case you have a workload that actually requires spatial data and your P data very common you can use one database for that you don't have to fragment your data across multiple cloud services and of course you get transactions you know you don't have to do anything you just say commit and you're done with these other this other cloud vendors you have to spread your data around all these different data sources dated cloud services if you want to do a transaction it's you're sort of out of luck maybe they have two phase commit across them they usually don't security you're again you're you're at the mercy of the weakest security of all these different databases you're storing your data in and your your fragment your data that's the worst part and so with our approach analysts get what they want they want to be able to run rich queries to understand the business and if data is fragmented all over the place they have to do lots of ETL to consolidate the data back together again so they can run queries okay you know that's that's not easy way to go so with Oracle you can keep the data consolidated so you can have a nice global consistent view of what's going on in the business of course the ops people also get what they want they want reliability they want security they want scalability and with Oracle you get that as well so this is my my first myth myth busting slide okay number two this is another thing you hear from developers though Oracle doesn't really fit into what we're trying to do with our cool new micro services applications and private clouds etc and here the answer is multi tenant you know I mentioned earlier that all of our cloud services autonomous database service Fusion SAS are all based on multi-tenancy and multi tenant is also the right technology to be using on-premises when your developers are running in private clouds you know whether cloud is based on VMS technology or it's on the latest kubernetes technologies what you want to do is give each one of your developers or organizations who are building a micro service their own private pluggable database so what would they tell the popular cloud vendors tell your developers you know for each one of these micro services you're building you should have a separate database because you want it to be totally isolated from all the other micro services and then you can upgrade things and patch things independently it's really great so would we tell our developers you know with Oracle that doesn't make any sense you don't want to give each micro servers its own Oracle database instead give a pluggable database to each micro service and that gives the developers sort of what they want they want isolation they want their own private schema they want to be other you know develop and enhance their application independent of the other micro services and so our advice to cut developers is use might use multi-tenant just like we use it in the cloud it works really well and ok and and then you get of course all the the scalability and reliability and security of the Oracle database ok end of my two myths section the next release 20c is in in development now it'll be coming out towards the end of the year early next year just like all the other annual releases in this release we've done a few things that are sort of interesting the first thing up there is native blockchain tables so you've all heard about blockchain right it's it's this technology that was built originally to support Bitcoin and what it does is it lets you have essentially an audit trail that's tamper proof and trusted etc using you know encryption techniques and you know one of the popular cloud vendors out there just created a new service they called Quantum ledger service and it basically is this blockchain technology where you again it's like it's the ability have a table of information that is tamper proof and so we said you know this is pretty stupid again it's like why do you need another cloud service just so you can have a blockchain table so with the Oracle 20c now when you create a table you can say I want this table to be a blockchain table just the it's just part of the DDL create table statement and when you do that it it uses all these standard technologies now that people originally build for Bitcoin for making a tamper proof table so this table looks and feels like a sequel table you can query it like any sequel table you can do inserts into it like any sequel table but you can't do deletes you can't do updates it's it's tamper proof it's encrypted we use these the same techniques for hashing the blocks so that new blocks are a function of the old blocks and we can detect tampering etc so that's new and 20 C and we think that you know about 95% of what people are thinking of doing with these distributed block chains can actually be done to this simple kind of table infrastructure auto ml this is another really important thing so again one of the people one of the things people are confused about about Oracle databases we actually do have built into Oracle database a very rich set of machine learning algorithms that we've been working on for 15 years now and the one of the problems in the machine learning space is that most analysts don't understand no machine learning algorithms and it's very you know and there's just a shortage of people that really understand this technology so what we're trying to do with the autumn auto ml is to make it possible that you know most business analysts will be able to use some of these machine learning algorithms to build models that they can use then for for predictive purposes so in 20 C you week now with Auto ml we will look at sort of your data you will give us sort of you know the predictions that you want you know here's a set of of target predictions here's the data this is what we think the right outcome should be it's sort of a training set for the machine learning algorithms and then we will look at what the right set of columns and your tables are that are to be used for the algorithms this is called feature selection in machine learning lingo for you and then we'll look at various algorithms that we have available and we'll actually figure out what the the right set of features are required in the right set of algorithms are unique can be used to get the right set of predictions and we'll do this automatically for you and then your your analyst can choose from different choices of algorithms we'll say this one looks best but you know you could use this this up this other one as well so that's what we call Auto ml we think this is really important because we really want to help you leverage the value of your data by letting more and more people be able to help capture the value it shouldn't just be a few data scientists who are often you know in the corner of some department but all the analysts should be empowered to be able to use this technology the last two features are also really important Jason of course is very popular and in 12c r1 we started supporting JSON documents which were textual documents you know character strings in 20 C we are now also supporting a binary representation of JSON which not surprisingly lets you query JSON documents traverse the complex documents do updates really fast so I don't have the numbers here but this is like 2 X 3 X 4 X faster for for update intensive work or traversal intensive Jason operations and then the last feature persistent memories store is really important I measured it earlier you know one of our goals of course is we want to build database software that's the leader 10 years from now and by far right and one of the big disruptive technologies that is now out there is this new persistent memory technology that Intel's when production with earlier this year Intel calls it opting and I'm sure most of you have never heard this thing but this is really important because all the standard algorithms for for databases were invented when discs were really slow and we did all this work to cache data in buffer caches to avoid iOS to disks and all those algorithms are now obsolete when you have persistent when you have a database that can fit fit in persistent memory and so we are working really heavily on two things we have evolutionary technology so in 20 C you can now store essentially a file in persistent memory and use that to store data in your database and when you do that I'll we will essentially use that directly like an init but you become sort of an in-memory database and we you have you know the persistent memory technology is relatively easy to use but when the system crashes you still have to deal with interesting failure conditions so it's not trivial to have a database support this kind of technology and in 20 C we have the first sort of evolution of Oracle to support persistent memory and then we are working longer-term on a project to actually reaaargh attacked a lot of the guts of the Oracle database for persistent memory we think this is one of these things that if you don't do it in five years you know you're gonna be a legacy product so that's a really important first step and persistent memory support okay so that's sort of a short summary of what we're doing with the database software as I mentioned all these innovations are in the software can run on Prem can run clouds front-end exudative etc exudate is evolving as well earlier this year we came out with the X 8 the X 8 is just like all the previous generations of exadata you get more compute performance you get more storage all at the same price and then we added one new thing a lot of customers were you know have data that gets cold after a while and they were thinking you know it's you know it's too expensive to store this data that I hardly ever query in my exudative storage so I'll move it off the Exadata to some lower-cost storage tier but when you do that of course you lose the ability to query the data so one of the things we did with the new x8 is we have a new tier of storage that you can buy called XT storage and we wanted to make the storage really cheap so when your data gets cold you can just move your data it's this storage tier that's on the Exadata you retain query ability but you sacrifice performance we to lower costs we just ripped all the flash memory out of the the Exadata and we also made it possible that although the data is queryable if you don't want to use Exadata you know smarts can to offload your queries you can tell us that and then we won't require you to buy the exudative storage sort of software storage server software either so this is a really nice alternative for those of you who have lots of data in your exudate is some of that data is getting cold you can now store it at much lower cost with the xt storage the other new thing which we are announcing today and i don't want to take all the the glory here from the you know larry will be talking about this at is keynote this afternoon the x8m is a new version of XA that's coming out this month and it is quite a radical departure so there's two things new here number one persistent memory is now in embedded in the storage hierarchy of an exadata so transparently you will get the performance benefits of persistent memory and one of the things you're gonna get and it's in that little lower left-hand corner of the slide you're gonna get 19 microsecond iOS on this on this box so that's you know that's pretty fast the other thing we're doing here is that ethernet has basically caught up to InfiniBand it now supports our DMA which is the big one of the secret sauces of InfiniBand they let different servers in a cluster access the memory of other servers in the cluster and InfiniBand also now has the performance we have a hundred Gigabit Ethernet every bit as fast as what InfiniBand can do in fact faster even so this new x8m is going to be the first generation of Exadata that is based on ethernet now of course for those of you who have EXA data's that are based on InfiniBand we will continue you know for quite some time producing exudate is also with infinite bands so you can grow your existing Exadata configurations but for the new customers who are people that are deploying new exudate is the x8 m will be the future direction and you know i mentioned earlier that future in the future exudate is will be natively provisional on the cloud well you can see why that's going to be possible this this Exadata x8m is now built using ethernet ethernet of course is what we use for networking on the cloud and so it's a real prereq you know that's sort of the beginning of our move to an exit area that's completely provisional out of resources that are just under compute storage etc on a cloud okay and just in case you don't know how fast nineteen microseconds is we we put some charts here to compare you know what is the Exadata x8m performance versus what you get on on various clouds on AWS and azure and yeah it's it's really fast okay now this one I hear sometimes from customers you know there's this myth that exadata is a lock-in product so you know it's a bunch of Oracle hardware so it must be locking me into something right and one of the things people don't quite understand it's our fault we didn't communicate this really well is that when we create the Oracle database we have these developer api's you know sequel of course is the most important of all the developer API is we have with the database when we sort of factor the product for what's works in the cloud and what works what on Prem and what works on exadata the api's are sort of sacred we want to make sure that any application you build with our developer api's will run everywhere it will run on Prem it will run in an exudate out run in our cloud now of course Exadata has really great performance and it really scales well and it's really reliable so that application will run better you know on an Exadata but it will run if you take it off in Exadata you can rent it anywhere and so that's one of our big goals to not to keep XA to be an open product and no one contrasts of course you all know about IBM mainframes and how you know these applications that were written 30 years ago are completely locked in and you can never move them ever you know that is a true lock-in product Exadata is really a much more open technology and then of course the cloud vendors you know these guys like you know Amazon have all these proprietary services those proprietary databases I was talking about earlier you write your applications at any of those api's those applications run in one place they run in their cloud nowhere else and so I think one of the things you know you have to be aware of as we go into this cloud era is that you really want to think about portability you want to think about using open standard API s and for data management sequel is the okk standard API and that's really what you want to build your database applications around is open standards not proprietary api's okay last last topic cloud of course we are working on cloud and autonomous database is the big initiative we have there and again just the contrast what we do versus the other popular cloud vendors you know we have a self tuning self driving service they have still manual tuning services we have the Oracle database at our core and you get all the enterprise reliability and security and scalability that Oracle has they have open source databases that are really nice but you know don't compete with Oracle and then of course they have all these specialty databases we talked about earlier you know they have nine different specialty databases on Amazon each one lets you fragment your data and run you know run your graph stuff over here and your blockchain thing over here and your analytics over here we don't think that's a really good idea anyway what we are doing is you know a self-driving database we build in enterprise class security and reliability using an Oracle database technology and we have a converged engine or a multi model engine that really can do all kinds of different workloads for you you don't have to learn all kinds of different API is you don't have to use proprietary API you can just use industry standard sequel for all your data management this is the architecture sort of at a high level of autonomous database and we have two separate autonomous database services one is called autonomous data warehouse and that's what Pat was talking about in his experiments that he showed you earlier autonomous data warehouses for all your analytical use cases like a data warehouse data Mart data science data Lake etc and autonomous transaction processing is for everything else you know it can do pure transaction processing workloads it can do mixed workloads of transactions and analytics which is pretty much everything you do you know 80% of what people do and at the core all of these services are identical you know there's a there's a big Exadata infrastructure at the bottom we build an eight node rack cluster on the exa data's we use Oracle multi-tenant on top of that to share the resources across multiple use multiple organizations that provision a database and then for autonomous data warehouse we take advantage of the Oracle in-memory columnstore technology to make it possible for you to do incredibly fast scans we offload the column store scanning to the Exadata storage as well which is why we can get so much better performance like what Pat was showing you earlier and so and that's all done transparently you don't have to know anything you just say give me five CPUs of autonomous data warehouse and with two terabytes of storage you push a button and you get this highly tuned high-performance analytical engine for analytic type workloads for transaction processing workloads it's much more complicated world there ute you tend to use our standard row format and that's what we do by default with autonomous transaction processing and use standard indexes to make things go faster and that's a much more complicated environment to do self tuning and and I'll talk about how we do that coming up okay of course since we are now autonomous we are doing the work that used to be done by armies of people we still have have lots of people - but not armies we'd have small little you know platoons or something and so to do the work we have all kinds of algorithms at all the different tiers some of which are classical machine learning algorithms some are expert systems etc - to do this this autonomous capability to automate various things so the Exadata infrastructure of course is exudative so we've been automating the ability to detect hardware failures in Exadata for years we detect hung servers etc in the database server as part of this service we use something called the autonomous health framework which is also available on premises to gather telemetry from all the database and look for anomalies and try to predict you know upcoming failures that might be happening and then we do all the self tuning around this as well so this is a complete service where we do the whole thing we manage the infrastructure for you we manage the database for you we do the self tuning for you it's mostly all automated now of course there are all kinds of performance to neat features in the Oracle database that we haven't gotten to yet but you know over the next five years we will be making sure everything is automated from a self tuning standpoint we're now now in production for about a year and a half we've got lots of production customers and you know this is great well you can still read the logos that's good and I just wanted to drill down into a few of the customers that give you a flavor of what the different customers are doing so the way this slide works is the top row is autonomous transaction processing customers and the bottom row or the autonomous data warehouse customers and in the transaction processing space one of the things you're first gonna notice is like I've never heard of these companies who are they and so a lot of these customers are relatively smaller enterprises the big more mission-critical customers are still using x8 a cloud service and so here jesse is an example of a customer they were on they do software for warehouse management order processing etc they were on IBM's cloud they moved to autonomous database and as you can see they're getting you know much better performance with autonomous database and they got with their previous cloud mess tech is interesting they do manufacturing software for for helping companies optimize their you know manufacturing processes these guys are most interesting because they weren't there one of the early customers who are writing your applications on Azure and running their databases on autonomous database and using this you know integrated interconnect we announced recently with Microsoft around a shirred on the autonomous state warehouse side TaylorMade is an example of an enterprise company you know they make golf clubs you sure heard of them and they had an existing Oracle database on Prem they moved it to atw on the cloud we have a lot of customers doing that sort of thing Hertz BP lots of customers have lots of these data marts you just repoint the ETL at our cloud atw you repoint your bi tool at the cloud aw and you're in business it's very easy to move data marts to the cloud Wow haiya is one of the bigger media companies in Latin America they're based in Argentina they were using Tara data on Prem over the last year and a half they've moved everything off tera data onto a DW and get much lower net they actually get much better performance as much lower cost so that's that's a really nice story q MP is a good example of a startup we're getting a lot of startup companies who are discovering that atw is an incredibly disruptive technology there in the healthcare industry their existing competition of course has this established IT infrastructure that's very high cost they are able to go in and compete in the blood testing analytics market and be able to deliver results you know in ours that used to take weeks for their competitors to deliver so atw is a really powerful technology for disrupting markets okay we have two kinds of autonomous databases now serverless is the original service that we came out with a year and a half ago it lets everybody sort of share this infrastructure in a cloud in a multi testimony tenant fashion you can start at very low cost just give me one CPU one terabyte or scale-up from their dedicated is the new service we just introduced a couple months ago and this is for customers who are really big they don't want to just have one data warehouse or maybe two they want to have 50 databases or 100 databases and they want they're big enough now that they want to have the the equivalent of a private cloud in our public cloud so they I can actually allocate their own Exadata infrastructure that's just for them and in exchange for that they get some isolation that they can't get with the normal server list version of the service and the big thing they get is is more and more of the control that they're used to on Prem so for example if we want to do patching or we want to do upgrades they get to decide when that happens you know they can give us they have to give us a window of some sort but they have control on the server left side you know you're sharing this infrastructure with all your friends and neighbors from other companies you can't really control that so this is a service we think is where the the enterprise customers are going to want to go dedicated is a great way for them to go to get maximal isolation okay automatic indexing this is the the new expert system we came out with for doing self tuning it does basically what a smart DBA will do it learns all your sequel statements it remembers the sequel execution plans it remembers the runtimes of those plans and then as your workload evolves it looks for new cut new indexes to be created that can make your sequel statements go faster just like a good DBA would do it tastes to test it on the side looks for things looks for indexes that work well when it finds something it'll roll it into production and the first time you run any new sequel execution plan that uses that index we will compare the performance with the old plan and if it's worse well immediately go back to the old plan on the next execution so this solves one of these really important problems that people have had for years is like suddenly my sequel blurry query is going slower it's using some just using the wrong plan because I did an upgrade of some sort now that problem is gone right if it regret if any of these sequel execution plans regress it'll run one time and then we pulled back so this is a really important technology this is all sort of now part of autonomous transaction processing in the dedicated world and it'll be coming out on the thomas transaction processing server lists in fact I think we're in preview mode now there as well and with that we're near the end here and I wanted to talk a little bit about apex so how many of you actually know what apex is I'm just curious raise your hands oh I'd say about happy good so apex is this really one of our secret sauces of Oracle that the people in the know sort of understand with Apex you can build applications really fast without writing lots of code and this is really important you know the especially in the cloud world the competition really believes that you need high powered developers writing tens of thousands of lines of code to get anything to work on these clouds with apex those of you are in the note you can build applications with a hundred times fewer lines of code often no code at all you can develop things of course much much faster without if you don't have to write code and so first of all we're going to run a little video with some apex developers who are going to talk about their experience with apex [Music] we're doing more things quicker and they look better and I perform on any device data warehouse applications permitting applications survey applications all using application Express points an NES ap table take that data analyze it into an apex application and then create the scripts to move that over to the Oracle ERP cloud they're able to use it our application we built an apex to fulfill all the needs of therefore 12 in one application they don't have to go here to do toxicology this for reporting the body they don't have to worry about dealing with other information and providing reports the third parties without a fax it would have taken us [Music] okay and now to to give you a little proof of how fast a px can build applications I'd like to invite out Jenny sighs she's our VP of Product Management databasing show you Thanks thank you very much we'll build an app for you in a few minutes okay Thank You Mandy hopefully I won't I won't flop here so I'm actually going to be showing you how to build a complete application using Apex 19.2 which is our very latest version you could try it at try apex now calm and I've downloaded some airbnb datasets so this is this particular one is all the rental listings in San Francisco and what I'm going to be doing is just uploading the data using apex into a database in the cloud and then creating an application without writing any code the only typing that I will be doing is typing in the name of the table that I want to create using this data set alright so let's get started so first of all I need to sign in to this apex 19.2 and going into my workspace next I'm going to use the app builder wizard to basically walk through the flow so again I'm not typing any code I'm gonna go ahead and upload the CSV file the spreadsheet that I downloaded from Airbnb it has about 8,000 records so not quite the quintillion records that Pat talked about but enough to show you that there's some real data here Apex went ahead and sampled some of the rows in the spreadsheet and determined the column types automatically I'm going ahead and typing in the table name and just clicking a button to load data and again it's about 8,000 rows so it should take maybe a few seconds to load and it's writing into the database in the cloud all right so we've got 8,000 rows thereabouts next I'm going to go ahead and click a button to tell it to create an application and this will create basically a home page for us along with a sign on screen plus a dashboard a interactive report that allows me to edit the data plus the faceted search page which is the brand new feature with 19.2 apex all right so I'm gonna go ahead and and by the way it was smart enough to know that maybe I want to use the table name as my application name so it typed it in for me already and there we go wallah in less than a minute I have a full-blown database application written by apex for me and it's created some of the pages that you see at the bottom here let's go ahead and run this application as if we're in users so first thing that's going to prompt me to do is to basically type in my user name okay and I'm just using my developer username and signing in let's take a look at the interactive report which is also editable form okay so here I can actually sort the data and even do some basic query of the data I can click on a button to edit the record and save it into the database so we'll just pick one here and add a bit of data here of course I could also delete records from here and create new records all right now let's take a look at the dashboard apex was smart enough to say there are certain columns that may be of interest and it's chosen certain chart types for me from here I could drill down into specific rows by clicking on for example room type equals private room it shows me all the records of that type now let's look at the faceted search so what apex has done is using some special algorithms to choose the best facets to use based on distribution of data the number of distinct values and it also uses the sequel grouping set to clause to calculate the counts automatically for me for each of the values in the facets so I'm gonna go ahead and choose bayview notice I have 183 records in Bayview by selecting room type Bayview is now reduced to 72 records that are relevant let's put in some minimum nights that I want to stay so one two three nights click go and again further reduction of the available rows based on filtering of the reduction of narrowing of the search next I'm going to go ahead and choose a number of reviews because I'm guessing that the more reviews a rental house the more highly likely that people rented it and are happy because they're taking time to write a review hopefully they're positive reviews and here I found one that says cozy private suite with separate entrance for only $78 a night that's a bargain I'm taking it thank you very much ok so we're gonna finish up now just do a few more slides and we'll let you get on your way so one of the big announcement Larry's gonna make today at his keynote is something we're calling the free tier of our Oracle cloud and I'll just give you a little teaser here for the Oracle database part of the free tier the big thing we're going to do is we're gonna make autonomous database available to you for free not just free for 30-days or free for one year it's a free forever offering it's for developers so we're gonna give you sort of micro versions autonomy database you can do either atw or ATP we'll give you up to 20 gigabytes of data we'll give you sort of a portion of a CPU sometimes you'll get up to one CPU and you'll get the full functionality everything there apex sequel developer machine learning notebooks etc so that's that's big that's something to let all the developers get our technology really easily for free over the cloud data safe so Oracle of course autonomous database has by default all the data is encrypted we do the security patching online for you automatically data safe takes care of the stuff that we can't do automatically it helps you deal understand with the privileges that have been allocated to your users it does auditing and captures audit trails and looks for you know kind of anomalous behavior in your database for you it's a new free part of all the database cloud services that we have on the Oracle cloud autonomous database today is available on our public cloud and all of our various data centers around the world and coming soon sort of towards the end of this year early next year we will make autonomous database available at your data centers just like Exadata cloud a customer today can run in your data center you can run our exited cloud service you'll be on the run autonomous database service as well in your data center and form your own local cloud and then finally just from a vision standpoint I mentioned it earlier you know one of our big goals here is that we don't want you to have to have developers writing thousands of lines of code to get anything done in our cloud and so we are you know building on what we already have with autonomous database and things like apex and Auto machine learning to try to make your experience of your analysts and your data scientists really simple if they want to get value out of the data and so things like ETL are just going to be rolled into autonomous data warehouse you're the the vision is your analyst should just show us the data they want to deal with the transformations they want to make and we'll take care of all the movement of the data for them will do the caching for them etc we will have a strong metadata catalog offering and it's being announced this week actually that we will use for this service and then high-performance machine learning algorithms both in the database like I mentioned earlier and then we have open source algorithms that we call from our data science organization that will also be available here so anyway that's just sort of a visionary standpoint so this is what we talked about today we talked about what we're doing in the software space 19c is here all of you should be looking at moving your your your applications on Prem or your ISV applications in nineteen c20 C's coming soon some interesting stuff there exudative course is constantly being revved and excavated x8m is a quite a new radical departure for Oracle where we are frak sedated where we are now moving to an Ethernet based infrastructure with persistent memory and then of course we talked about autonomous database and with that I'd like to close and I'll invite you to come to some talks we have you know later this week where you can learn more details about all the things we talked about today ok thanks for coming [Applause]

Channel: Oracle

Views: 7,010

Keywords: Oracle OpenWorld 2019, Oracle OpenWorld, Oracle Open World, OOW2019, OOW, Keynote, General Sessions, Andrew Mendelsohn, Database, Database Server Technology, Database Server Technologies, Pat Sullivan, Managing Director, Accenture, Autonomous Data Management, Autonomous, Data Management, Data, Operational Workloads, Analytical Workloads, On Premise, Cloud, Database Cloud, Autonomous Database, Database Development, Cloud Platform, Database Cloud Service, Analytics, 40151159, 2019

Id: QN7hFOpSSUI

Length: 77min 41sec (4661 seconds)

Published: Wed Sep 25 2019