How Autonomous database and AI change the work of database administrators

Oracle Academy Tech Chat

Artificial intelligence or AI is everywhere from movies to healthcare. In this episode, Tyra Crockett Peirce speaks with Oracle database expert Ace Director, Jim Czuprynski, about how AI is changing the jobs of database administrators and the skills that students need to learn if they want to pursue a career as a DBA.

--------------------------------------------------------

Episode Transcript:

00;00;09;03 - 00;00;27;21 Welcome to the Oracle Academy Tech Chat. This podcast provides educators and students in-depth discussions with thought leaders around computer science, cloud technologies and software design to help students on their journey to becoming industry ready technology leaders of the future. Let's get started. 00;00;27;29 - 00;00;57;02 Welcome to Oracle Academy Tech Chat, where we discuss how Oracle Academy helps prepare our next generation's workforce. I'm your host, Tyra Crockett. Artificial Intelligence, or A.I., is everywhere, from the movies to health care. In this episode, I speak with Oracle database expert and ace director Jim Czuprynski about how it is changing the jobs of database administrators and the skills that students need to learn if they want to pursue a career as a DBA. 00;00;57;14 - 00;01;24;03 Jim Czuprynski has been an Oracle DBA for 20 plus years. During that time, he's taught 2000 plus students in the finer points of Oracle database administration. As an Oracle University instructor coauthored four books on Oracle technology and has spoken dozens of times at user group conferences around the world. During that time, he's seen plenty of dramatic changes to the DBA role, especially after autonomous database was introduced. 00;01;24;14 - 00;01;25;09 Welcome, Jim. 00;01;25;19 - 00;01;27;05 Thank you, Tyra. It's great to be here. 00;01;27;11 - 00;01;32;15 To go ahead and get started off, can you give me a little bit about your background and your current job role? 00;01;32;26 - 00;01;56;08 Absolutely. These days, I'm more of a technology advocate, but to be honest, I'm still at heart an Oracle DBA, as I have been for over 20 years. I go back all the weight. Oracle 81723. So as you might imagine, I've seen quite a few changes in the Oracle database and the hardware that it runs on over that time. 00;01;56;18 - 00;02;20;18 These days I'm actually kind of back into some of the really neat features that are built into just about every Oracle database. Especially Oracle 23. See our latest release that's almost ready for general availability. And most platforms. And I really enjoy delving into things like graph and spatial and machine learning and analytics. 00;02;21;18 - 00;02;31;27 So my next question is kind of a little bit along those lines. The Oracle Autonomous Database was launched a few years ago. How does it differ from previous versions? 00;02;32;15 - 00;02;59;12 Wow. The autonomous database, or most of us call it ADB for short. It has features that are already present in the most stable release of the Oracle database. That's typically on premises right inside an organization's local data center or maybe even an on premises data center. But the big difference about autonomous database, first off, is that it runs inside ORCL cloud infrastructure. 00;02;59;19 - 00;03;27;25 And best of all, it runs on, in my opinion, the best hardware ever built for Oracle database workloads. Oracle Exadata, which is an extremely robust environment. It eliminates a lot of performance bottlenecks that are inherent in traditional on premises database hardware and I speak from experience. I built servers back in the day to run Oracle 8 and 9 even 10. 00;03;28;19 - 00;03;35;22 It's not trivial to build a robust database working environment. 00;03;36;09 - 00;03;44;00 So let's kind of go on to AI a little bit. So how is a I used in an autonomous database or ADB? 00;03;44;27 - 00;04;20;19 That's a great question. EDB uses A.I. in many different forms underneath the covers, and it really automates a lot of tasks that a typical DBA doesn't necessarily have the bandwidth to perform everything from something called automatic indexing. We'll talk about that a little bit later, I hope. Things like automatic scaling of CPU's based on demand that it observes automated database patching, and even the ability to automatically identify. 00;04;20;19 - 00;04;47;01 And this is really cool, a poorly executing SQL statement and essentially quarantined them so that somebody from the DevOps slash DB 18 can scrutinize them and go, Why are these things taking up so much bandwidth as the next release? Oracle 23 C rolls out for ADP. There are several new AI controlled features, including and these are really cool. 00;04;47;10 - 00;05;17;29 Automatic transaction rollback, for example, a transaction might be holding a roll lock and blocking a higher priority transaction. You can actually configure that so that that will happen within boundaries automatically. Another thing that's really neat is automatic generation of materialized view, and I worked with projects even in the last 12 to 18 months where materialized views are a tricky thing to build out. 00;05;17;29 - 00;05;50;01 You really have to kind of know what's going on when you build them. Now, in latest release, that's going to happen automatically. By the way, these features I'm talking about, you do have a choice as a DBA to go in there and if necessary, turn them off. I get that question a lot. So, you know, you can definitely deactivate them or essentially turn them down and make them advisors temporarily and then decide whether or not to turn them on in full for your environment. 00;05;51;04 - 00;06;29;04 Perhaps the most intriguing new ADB feature is retrieval augmented generation or RG or RAG for short. RAG employs several large language models or limbs that translate a request for information. For example, find the top five customers within a mere based on their average percent increase in sales for the last six months and simply by prefacing that with the words select a I, I hope you're sitting down as you're listening to this, it literally will build the skill statement for you. 00;06;29;18 - 00;06;49;03 You can even say, just show me what the SQL statement looks like, that you just built your ADP and it'll show it to you. So again, one of the ways people are using A.I. these days, right, is more of a helper. You know, don't don't write it for me, but show me what it would look like if you wrote something like this. 00;06;49;12 - 00;06;55;14 And now you can do that with Rack. That's to me, mind. And amazing. 00;06;56;07 - 00;07;11;16 Then kind of one of the things that always comes up is if we're using AI for all these other things, like, you know, writing SQL statements or like helping all in patch or data, how does it impact our security for our database? 00;07;12;13 - 00;07;35;27 Well, another great question. Like I said, I've been doing it for 20 years as a DBA for two years in all different angles. Right. And one of the first things that if I'm on a DevOps team and someone proposes something, the first thing I'd like to ask is, yeah, great idea. Sounds really radical. Sounds cool. How secure is that? 00;07;35;27 - 00;08;01;02 Right. I want to think about not just what's going to happen at that database level, but all the way up through that seven layer salad, if you will, all the way out to eventually our firewall. Right. Is this going to be solid and secure for application access layers? What if I have a special purpose database inside of my my firewall? 00;08;01;03 - 00;08;32;20 Right. Maybe something specific to spacial or graph or mapping or whatever it is I like to think about How easy is it for a bad actor to do something as simple as SQL injection to gain access to either the data insider databases or even the infrastructure. I mean, a day doesn't go by that we don't hear about some sort of hack or some sort of ransomware attack that has happened because of someone ignoring a business risk. 00;08;33;01 - 00;09;00;27 So one of the things I like about ADB and Rack is that I'm eliminating several possible attack vectors. Right. One of the things that I'm doing with using select a I, and then a question is I'm letting before called database write the query. It's a lot less likely that hopefully at least I haven't seen it do it yet. 00;09;00;27 - 00;09;40;03 I've been playing with it where Rand might go ahead and write some kind of weird thing that would do SQL injection inside my database. I haven't seen that happen and I don't anticipate that would happen. So I really like that aspect of API being implemented and if you will, really within the environs of the database itself. Also, one of the things that is evident to me, having talked to people about all the other kinds of databases, even open source databases, Oracle is well known within the industry as the most secure enterprise database. 00;09;40;20 - 00;10;08;13 That means that even if someone were to gain some sort of access, I have excellent tools at my disposal that will identify that access immediately. And this is a big advantage, especially for the concept of what we call converged database, which I've been talking with several other ace directors and other DBAs and developers for almost three or four years now. 00;10;09;01 - 00;10;44;02 The whole idea is more likely than not if you're building a graph. DB solution heck, even a vector database solution with some of the new things that are coming with the next release of Oracle and maybe even a spatial solution, all these things are already built in to my database environment. If I'm doing a spatial solution, a graph data property graph solution, a machine learning solution, there are dozens of machine learning algorithms built into an Oracle database. 00;10;44;10 - 00;11;05;28 Why would I want to take a risk of having something outside my environs that compose my database and even my computing infrastructure? So from my perspective, a I, especially an oracle, is going to be something that presents a lot fewer security challenges. 00;11;06;22 - 00;11;25;19 That is very comforting to know that. But then one of the things that I hear about AI is moving to ADP, and we've been working with it at least on our team for the past three plus years now has an autonomous database. And then I change the work of a database administrator. 00;11;26;04 - 00;12;00;13 Hmm. Great question. Again, let me go to one example that I kind of alluded to earlier for ADB. There's really two subsets, if you will, main subsets, one called Autonomous Data Warehouse, which, as you might guess, is aimed at data warehousing and data lakes and more very driven type activity. There's also an autonomous transaction processing ADB flavor. I have one up and running right now in the Oracle Cloud infrastructure environment for three years. 00;12;01;08 - 00;12;45;05 I've never had to mess around with all of the things that I would have if I was not running autonomous. So one thing I alluded to earlier, this feature called automatic indexing, it's really specific to autonomous transaction processing or ATP ADB instances. I actually did a test study with the transaction processing counsel's TPC e workload and built out quite a large database with that and then ran the query workload against my database with no indexes, no secondary indexes. 00;12;45;05 - 00;13;11;29 Right. To be more specific, primary key indexes and other things for referential integrity were there, but no indexes that would normally help a query perform better. And it was so neat that automatic indexing, once I turn it on within a very brief period of time, within about I believe 45 minutes had identified critical indices that needed to be added to my database. 00;13;11;29 - 00;13;38;09 And in one case, I remember I had one query that actually performed its improved performance, improved this performance by 1 million times by adding a single index. Now, if you're indeed one of the reasons that's important is we spend a lot of time hearing complaints from our DevOps team or even our end users going, Wow, it's running slow. 00;13;38;09 - 00;14;05;23 Why is it running slow? And one of the typical ways of solving that is slapping index on this particular table, right, to make it run faster. That could have amazingly deleterious effects on transaction processing. So one of the cool things where AI is actually being used underneath the commerce is this automatic indexing tool. It adds only the indexes that need to be index added to the database tables, which is mind blowing. 00;14;05;23 - 00;14;32;01 To me. It was really cool to see another one and this is relatively recent. We've had tools like this since early TNG, which is like 1518 years ago. There was a concept of automatic sequel tuning and this has been around for a quite a bit of time, but in more recent releases, yeah, like ten years ago, this has been around actually, and it's gotten better and better. 00;14;32;10 - 00;14;59;19 Automatic SQL tuning will look at a particular query or DML statement, even and go, No, there's a better execution plan that you should be using and it can either just give you the advice as a DBA and you can go evaluate it, decide to implement it or not or actually and this is really cool what you build confidence in it, right? 00;14;59;19 - 00;15;24;13 I said many of these things are actionable. You can turn it on when you have a certain level of confidence. So the stuff that I used to worry about as a DBA back 20 years ago, heck, even ten years ago, because of the learning algorithms and the AI that are underneath the covers of ADP, I don't need to worry about that stuff I haven't looked at. 00;15;24;13 - 00;15;46;19 And for my DBA colleagues, you'll know what I mean. I haven't looked at it a lot for my ETP database in literally three years because it just was fine. And if you were what I would like to call a helicopter DTA and always hovering over your database, you don't have to do that anymore. So what can you do instead? 00;15;46;21 - 00;16;16;18 Go help your team, build better systems, get in front of your DevOps, pick up efforts to build better systems, worry about things like better data modeling, worry about more efficiently written code from the start instead of, Oh my gosh, that's running terribly and then patching it later making that SQL statement or even application one better. So it's been quite revolutionary, to be honest from my perspective. 00;16;17;10 - 00;16;29;13 So then I'm going to transition to my next question, which is, if I'm a student, what are some of the skills that a student should learn if they're planning on working in a department database? 00;16;29;13 - 00;16;53;22 The one that I would advise for all students who are going to work anywhere near any kind of database these days, structured query language as well, Right? It's funny, I talk with younger people and their eyes kind of roll when I say that the same way my eyes would roll when someone said you should really learn assembler never worked in assembler, never worked in C courses like that was COBOL, right? 00;16;54;05 - 00;17;20;18 But the thing is, that school sequel is going to be around for a long time. Even for the Notorious not only or no SQL databases, you really do need to understand and if not, learn how to write SQL statements because you really need to understand what's going on underneath the covers. And that's just for manipulating and retrieving data from or inside your database. 00;17;20;28 - 00;17;51;14 And then the next thing I think that's really important is to think about how data relates to itself. As we talk about these days, especially things like graph databases where we're really mapping out not so much the fact that this invoice has a bunch of line items, right, but that this customer, because they purchased this item, influenced another customer who influenced another customer, who influenced another customer to purchase that same item. 00;17;51;23 - 00;18;16;17 So things like property graphs and vector databases, these things are really coming to the fore. So learn a skill for sure, understand it so that when you see it, you can interpret it and understand what's going on because it's simply not going to go away. There have been movements to try to kill it, but I don't know how much longer it's going to be around. 00;18;16;17 - 00;18;22;21 But from what I'm seeing, it's going to be around for a long time. Why not learn it? Why not master it? 00;18;23;07 - 00;18;42;29 I do want to add one follow up question to that is what about with Chat GTP and how it can write cycle statements? Should I still learn how to do SQL if I've got like my modern language processing skills or some of these language tools coming up, is it still really important to learn like the basics of sequel? 00;18;44;02 - 00;19;11;00 Absolutely. You know, sequels really simple that it's elements. It's about 20 verbs, and that's really it's super simple. And once you understand it and more importantly, that you understand it's set based, right, which is different. It's the Venn diagram stuff that all of us probably struggle from or through in grade school and even maybe even high school. Right? 00;19;11;04 - 00;19;38;20 But set theory is at the heart of effective processing of data, right? So it really comes in handy, number one, because that's one of the most effective ways to handle relationships between data. And secondly, I really hate to tell you this, sooner or later somebody's going to say, C'mere here, here's some SQL from 20 years ago, I'll fix it. 00;19;38;23 - 00;20;06;05 I'll figure out what's wrong with it and enhance it. You're going to end up having to look at it anyway minus will learn it now. Right. And as long as we're talking about Chartbeat and these other things, like I mentioned, us select a AI statement that you can use inside Adbe right now. You can use that if you want to learn how to build really good sequel statements, right? 00;20;06;05 - 00;20;30;18 So you can use it, if you will, as sort of a modeling tool so that you can see how a really good optimizer, a really good tool, builds well-written sequel statements. So I'm not saying throw it away, I'm just saying make sure you're using the right flavor of cheap, which by the way, the air to a rack is sort of built on. 00;20;30;18 - 00;20;42;28 Right? A good elements of it, many elements of it are built on, I believe Chachi beat 3.5. So you're not throwing away anything you were learning there. It's more synergistic learning, if anything. 00;20;43;10 - 00;20;53;25 That that was actually really helpful. And so now to my very last question, If you could give one piece of advice to faculty, your students, what would it be? 00;20;54;03 - 00;21;15;27 The advice that I didn't listen to when I was younger is the advice I'd like to give you. I spent my time believing that if I can master the code, that's all I'll need to be successful. And boy, was I an idiot because I had people, my mentor would go, You know, Jim, you're a really good coder. You're really good at what you do. 00;21;15;27 - 00;21;38;12 But to be frank, you're really upset. People don't interface with them very well. You're a little you come off as arrogant, you come off as mean. You don't seem to really care about people as much. You should really concentrate on making those things out, making those things better. What do you know? Oh, my gosh. I wish I had spent time doing that. 00;21;38;26 - 00;22;13;26 I really believe that. Secondly, what I found and this is an interesting piece, I mentor quite a few younger folks, even folks that are leaving the U.S. military in the Navy right now. My biggest piece of advice is network, network, network. You've got to build a network of humans because to be honest, almost every single opportunity that I've had, whether it's been within a company or maybe even a new job or a new opportunity, has come from 90% of the time. 00;22;14;04 - 00;22;44;04 Someone I knew it in my network. I wish I had built that network out much sooner. Again, you have a lot better tools these days. You've got social media and LinkedIn and all other kinds of tools. We didn't really have some of those 40 years ago, but there's no excuse not to build that network out. Seriously, that's twice as important, soft skills wise, than even learning how to write sequel. 00;22;44;12 - 00;23;00;10 I would actually say I agree with that. My very first job out of college was actually a neighbor that I knew that I, in fact said I was sitting next to him at a wedding reception and we were talking about how I just graduated college. I was looking for a job and he's like, Oh, I've got a job. 00;23;00;22 - 00;23;27;20 I might come apply for it. And that's that's exactly how I got my very first job out of college was through networking. So I wholeheartedly espouse that as well. So a big thank you to Jim for speaking with me on an autonomous database and the changing roles of a DBA. To learn more about Oracle Academy and our resources, visit Academy dot Oracle dot com and subscribe to our podcast. 00;23;27;20 - 00;23;28;16 Thanks for listening. 00;23;29;04 - 00;23;37;19 That wraps up this episode. Thanks for listening and stay tuned for the next Oracle Academy Tech Chat podcast.