IDUG Db2 Table Talk
IDUG Db2 Table Talk
Chemistry, Coding, and Collaboration - with Jørn Thyssen
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
From quantum chemistry to Db2, Jørn Thyssen’s journey is anything but typical. What started with a PhD and a few too many acid-damaged lab coats led to early work in neural networks, and eventually to Db2 via a “weird platform” called the mainframe.
His story shows there’s no single path into our community, and his advice is simple: learn by doing, even when things break. Along the way, he shares insights on where Db2 is headed next.
Learn more about IDUG, the International Db2 User Group, at www.idug.org.
Hi everyone. Thanks for joining us for this series of iDug Podcasts, where we speak to some great people in the DB2 community. I'm Marcus Gavage.
SPEAKER_03And I'm Julia Carter, and we represent the iDug Content Committee. We hope you enjoy today's edition.
SPEAKER_01Joining us in the studio today, we have Juren Tissen, who's the Senior Technical Staff Member at Rocket Software. Welcome.
SPEAKER_03Welcome.
SPEAKER_00Yeah, thank you for having me. I'm a longtime listener of your of your podcast, so it's uh nice to be here.
SPEAKER_03Laura, great to have you. And are you our first Rocket person? I think you might be.
SPEAKER_00Yeah, that's probably well, you had uh Anthony a while ago. He he had a brief stint at Rocket Software.
SPEAKER_03That's true, yeah. Yeah. But current employee, I think you're our our first.
SPEAKER_00Yeah, thank you.
SPEAKER_03Laura, it's good to have you here, and we're excited to hear about your life with and without DB2. So where should we start? Right at the beginning, how did you get into DBT?
SPEAKER_00Well, yeah, that's kind of an interesting story. Um I I do have like a different background than than most people. Um I have actually I have a PhD in uh relativistic quantum chemistry.
SPEAKER_03Yeah, I had to Google that because I read it and I thought I have no idea what it is about.
SPEAKER_01Okay, what is that?
SPEAKER_00Uh well well, quantum mechanics is quantum mechanics, but then sort of the chemistry part of it is really applying it to um chemistry or chemical uh problems, and then the relativistic part is really applying the the um the special relativistic theory to that. So um right now when we are recording, there's been a lot of discussion about gold and silver, like last week, if you follow that. And if if you know the periodic system with chemistry, that you know that gold and silver or silver is uh you know they're in the same column of the periodic system. So you would expect them to have similar properties, chemical properties, and so on. But you wouldn't you know that silver is silver colored and gold is gold-colored. And the reason why they are a different coloring is because of the relativistic uh you know effects on on the atom because the electrons are moving close to the speed of light. Uh so that's kind of the very short explanation of that. Uh.
SPEAKER_03And didn't you do didn't you do a degree in chemistry as well?
SPEAKER_00Yeah, so I I well I have a master's degree in chemistry first and then the PhD subsequently.
SPEAKER_03Um, chemistry was your passion.
SPEAKER_00Um it started out being chemistry, but then when I had like experimental chemistry in the lab, uh I would normally ruin for like thousands of pounds of equipment every time. And my lab coat was like full of uh holes from when I spilled you know assets and whatnot. So at some point, I I don't know if it's if it was me or somebody else sort of pointed me towards a theoretical chemistry rather than the experimental chemistry.
SPEAKER_03You sound like a stereotypical mad professor then.
SPEAKER_00Uh yes, yeah. My lab coat was like green and blue and red, and and yeah, that was not not good. Yeah. So it is chemistry, but I kind of found that uh theoretical chemistry is like one-third math, one-third computer science, and then one-third physics, so it was very little chemistry. Uh and most of my time I spent coding uh Fortran 77 uh programs, sort of implementing the the physical formulas in in Fortran 77.
SPEAKER_02Oh wow.
SPEAKER_00Running on you know, cray um machines and silicon graphics and IBM AIX and so forth. So that was how I got into sort of the IT uh IT part of it. Then while I was uh studying at university, I started playing backgammon and uh I joined the open source backgammon project, um, GNU Backgammon um that were writing an open source backgammon uh program. This was written in C, so that was how I sort of became proficient in C. As a side note, uh GNU backgammon was based, the engine was based on neural networks. So this was very early days. The the the brain of GNU Beggammon had a an N well today LLM type thing with I think like 10,000 parameters. No, today's LLMs or neural net, they have like 100 billion or one trillion parameters. So it was it was in the very early days. Uh so I worked with neural nets before it became cool. Well done.
SPEAKER_03Some would argue that it's always been cool, but yeah, yeah, yeah.
SPEAKER_01So you got your PhD in relativistic quantum chemistry.
SPEAKER_00Yeah.
SPEAKER_01Say what?
SPEAKER_00Well, so so I'd done a lot of the coding in C, uh both at university and then uh in that open source project. And then IBM was looking for C programmers. They did mention some weird platform in the uh job advert, but I did a platform I knew nothing about. But I knew C, so you know how hard can it be?
SPEAKER_03And that weird platform was uh MVM.
SPEAKER_00Yeah, it was IBM mainframe. Uh so I I knew nothing about the mainframe then. Um but I started in IBM as uh like a junior developer or trainee or whatever you call it back in in 2001, writing C programs on this weird platform to make it even weirder. Sort of the transaction manager was IMS GM. Uh but the the database was was db2. Um but don't tend anyone that I know anything about IMS.
SPEAKER_01Don't worry, your secret safe with us.
SPEAKER_00So so that was kind of how I uh started on on mainframe, really as an application developer. This was like IBM's internal IT, we were creating uh internal applications, but yeah, they were running on mainframe, IMS, TM, and DB2 as the database engine. Then over time it within that application, I became the application DBA. So that was kind of how I made it into sort of more the the DBA or or more DB2 side uh of things. Then fast forward a number of years, I think it was late 2009, uh, my manager said that they were looking for a DB2 expert in the in the in the IBM sales organization, like a pre-sales specialist. And so I I moved over uh yeah, probably late 2009 or early 2010 and and became a pre-sales specialist. So kind of a switch because I had I had the application background rather than sort of a system system DBA background. So I did have to learn a lot of things. Luckily, I got one of those uh CPDTs, like the mainframe emulators, and I that was uh I I I can hear right now that there's some discussion about IBM removing it, but uh for me back then it was really a good source of of education because you had your own system, and if you screw something up, you would just sort of start from scratch and and try again. So that's that's kind of how I learned initially sort of this the system side of things. So I I think I ended up knowing too much about SMS and REGF and all kinds of other things mainframe related from from that CBDT. Uh because you when you run the CBDT for half an hour, then you know SMF runs full or Jess runs full or whatever. So you kind of had to learn a lot of uh systems programming to to keep it running.
SPEAKER_01That's when you become dangerous.
SPEAKER_00Yeah. Yeah. Yeah, I I know enough to be dangerous. And that that's also kind of how I came into the idoc world. So I was working as a pre-sale specialist, and a lot of my customers went to the idoc conferences in Europe. And so I I started going a well, probably 2010 was probably the first one. And I I can have no idea where it was, but I attended all the iDocs probably since then, maybe missed a single one in Europe at least. Um so initially I I I I went as a participant, and for me it was like a learning experience uh because I I was new sort of in the system side of DB2 and the engine side. So I went to all the the presentations with the iBemers back then and learned a lot about the engine. Um I I don't hope Bart Stigman's this listing, but I remember probably one of the first years I had like a whole afternoon on SMF data or you know, IFC IDs in DB2, and that was probably a bit over my head at at that time, so I think I almost fell asleep.
SPEAKER_03A few hours of that, yeah, I can I can understand.
SPEAKER_00Yeah. Um, but uh it was really good education for me uh back then. Going back to my open source sort of um uh contributions, that was kind of uh where I started with sort of um you know community contribution or what you would call it. Uh we had a mailing list and I would participate, and that's how I ended up sort of coding on the project as well. Back then, at some point, I I joined the IDoc list um and sort of at some point I can't remember exactly when started contributing and answering questions or participating in the um in the discussions, etc. So obviously, I I think that that's also good education, sort of to participate in in the discussions. You learn something, you know. If you write something, it's wrong, somebody will correct you. As long as it's constructive, then I think that's that's very good as well.
SPEAKER_03Yeah, and it's it's good that you bring that up because whenever I search for something and it's the the DB2L as it used to be, your name is always there in the conversation. So you're always contributing. And it must take quite a lot of time to be constantly checking and and adding to that, but your contributions are really valuable.
SPEAKER_00Yeah, thank you for that. Um well, yes, it is time, but I I I still see a lot of it as education because if if somebody is asking that question on the IDOC list, you know, I might get the same question from a customer at some point. And so it's in a lot of cases it is valuable. Or it might be something that a customer already asked me, so I already know the answer. So it's just a matter of writing it.
SPEAKER_02Sharing it, yeah.
SPEAKER_00Yeah, yeah. So but I I think in general it's it's very valuable to participate in communities like that, um, you know, for yourself but also for the the ones that wrote the question.
SPEAKER_03Um so yeah, and you speak now. I mean, you say you started in 2010 and you were there to learn, but now you speak in every conference all around the world, and your sessions are always really well attended, and I think you get a lot of engagement, and I think people take a lot from your sessions because of the way, the unique way that you can explain things so that everybody gets it.
SPEAKER_00Yeah, thank you for that. Uh yeah, so as I mentioned, I started participating in iDoc, and then at some point I can't remember when I started submitting presentations. I I always kind of had an interest in sort of new stuff that's come along. So that's has frequently been the topic of my presentations. Or it's somebody that's a customer, something that a customer asked about, and I had to do a lot of research and then I kind of make that into a presentation. So I I presented on sort of newer things like your Python, pipelines, Ansible, and and so on and so on. Um so that's you know, from a personal point of view, it's it's always nice to learn something new, sort of keep you fresh. Uh so I like to spend some time on on learning new topics that that come along in in the in the db2 area or sort of around it. And with Python, I actually learned Python back at at university as well, many, many years ago. Python is quite an old language as well. But it's it's it's becoming more and more popular.
SPEAKER_03You were doing it at university, so that was a few years ago.
SPEAKER_00Um late nineties, yeah.
SPEAKER_03So yeah, I didn't realize it would be around that long.
SPEAKER_01Python was written in around 1990, something like that. Uh the early 90s.
SPEAKER_00Yeah. Yeah. So because I had an interest in IT, I I actually took some parts of the sort of the computer science curriculum at university, even though I was a chemist. Um, and and we a lot of the um they were using Python for a lot of the courses there, uh, like algorithm theory and so on. Uh I I did one on assembler programming and and I did one where we use C, so I kind of that's where I learned some of those programming languages. Um then Python was kind of dormant for me for many years, and then it kind of because it sort of became useful in the mainframe world, it's kind of where I picked up the interest again. So I have like a 15-year gap, Python gap.
SPEAKER_02But luckily, yeah.
SPEAKER_00Um yeah, it's it's becoming more popular now. Um uh you know, with younger people that do not necessarily want to or have to learn REX. Now for some things Python is a is a nice replacement. Um I I like it a lot. It's very it's very flexible, and there's a lot of libraries of modules you can you can pick and you know you know, passing JSON or YAML is like one line of code. I I would hate to write that in REST.
SPEAKER_01Yeah, I wouldn't like to do that in REX. I'd like to take this opportunity to thank you personally for helping me with Python and trying to get it working, talking to DB2 using the uh IBM DB2 libraries on the mainframe and talking to the mainframe. Because I've just recently been mucking around with Python. And it's uh yeah, you've been very very helpful. So thank you for that. Everything's up and running and working now.
SPEAKER_00Yeah, it it's a bit complicated and and it's somewhat different from running it on, let's say, your laptop. Yeah, just like yesterday that was I had a question from another customer. Came internally that somebody had asked one of our about one of our products and Python and IBMDB. So there's there's a lot of interest in that uh and and Python in general. Maybe I should sort of dust off my original Python and IBMDB presentation and and resubmit. Uh I think it was recorded for iDoc North America and it's out there on YouTube and it's I would say 90% still valid, but there's a few things and a few improvements since then.
SPEAKER_03So yeah, sounds good. Yeah, that on the agenda for Porto, maybe.
SPEAKER_00Yeah, yeah. Maybe I maybe I'll do that.
SPEAKER_03Do you think that the the modernization piece that we're all faced with at the moment is one of the bigger challenges of your customers at the moment in the fact that there are people that want to do it that are almost pushing for it, but there is quite maybe resistance to to actually go with it.
SPEAKER_00Yeah, I think that's exactly right. I I think a lot of customers struggle just to keep the lights on on their shop and they don't really have time to look at other things. But uh on the other hand, if you want to attract um new uh younger people to the platform, I I think we need to have some alternatives to to ISPF and and batch. So I I don't see ISPF going away anytime soon. Not in not in my lifetime, maybe, but uh I I do see like a a shift and an interest uh towards more modern interfaces. But of course, uh a lot of the products that we we all work on in the db2 tool space, they have no 20, 30 years of legacy and it's 1,000 functions, and it's really, really hard to make every single function and all of the options available in a modern UI. So that's yeah, but that is a challenge that's going to be a challenge for the future.
SPEAKER_01And Zoe, as part of the open mainframe project, is making uh a good dent in that respect.
SPEAKER_00Yeah, yeah, Zoe and there are many other things uh that are trying to you know add those modern interfaces. So yeah, at some point you know we we need to have everything available in a different way. And it can be web UIs for some features, or the features is more API that you can consume from, say, a Python program, or you can write your own sort of integrations, might be CLIs, you know, all kinds of interfaces that allow you to perform things. You can all you're all starting to see you know Python libraries for some features. IBMDB is obviously one thing, but now there's also SDSF available as a Python library, so you kind of get more and more functionality available, you know, some modern way, you know, API libraries or whatever. Yeah. So you can you can build your own integration. Uh I think it's also good. Maybe in the past, customers would either be, you know, BMC or Broadcom or IBM, but with APIs, you can kind of mix and match and get the best of everything. Even though we might prefer to have one specific stack, uh, I think it's still it's valuable for customers to be able to mix and match and even have like open source components at as part of that.
SPEAKER_03Yeah, that's a really good point. It does add that flexibility, doesn't it, potentially to to pick and choose the best or the most applicable bits to what you're trying to achieve or the problems you're trying to resolve.
SPEAKER_02Yeah.
SPEAKER_01I have seen uh threads on community forums where people have been saying, yeah, but you can't write ISPF panels in in Python, but you can in REX.
SPEAKER_00Well, then the question is if you need ISPF panels, you can easily create text-based UIs in Python. Then you would run it in in uh in ZOS Unix instead. But yes, it would not be ISPF, uh, it would be a different text UI. But it it may or may not be possible. Maybe customers or some users prefer web UIs and others they prefer more more CLI-based uh operations. So I I don't know if I I buy that that there's no ISPF in Python. I I don't I don't think it's necessary.
SPEAKER_01Yeah.
SPEAKER_00It's the change of mindset though that hey, I have this nice ISPF panel, and now I want to have the same, but with Python at the back, and well, no, sorry. You'd either have to write a web UI or text uh UI. I I have great success. Uh you call it wipe coding. Uh use Copilot for Python. It's it's very good at Python. Uh and I I just say, hey, give me a web UI, this and this and this, and it'll write the code for me. So you can you can you can quickly create uh you know not production level code, but at least some prototypes to see how it how it might look like and and work. So as a as a former developer that had like a 15-year gap in my development, I did not do much development, then having tools like that to help me is is very, very useful, I find. Yeah. I don't know, maybe I'm 90% vibe coding, so I just ask Copilot and then I'm making small modifications to the to the resulting code.
SPEAKER_01I've played around with um vibe coding. Uh can't stand the term because it goes against decades of uh software engineering training. Uh but just uh to just to talk to something which has got full access to uh everyone's libraries and everybody's bad code. Uh I have had good experience with it, and it is it is good fun. And I've churned out some really good Python programs which access db2.
SPEAKER_00Yeah, exactly. Um I I had a customer request come in, and they they from one of the tools we do for IBM, they wanted to have some an a new report or a new layout of an existing report, and and they wanted to save it in a db2 table. And I I I created a prompt in Copilot and it spit out 200 lines of Python code, and I had to make two modifications. Um the the DDL was embedded and it had like a syntax error in the DDL, and then going back to the IBM DB module, it was using like a distributed connection string to DB2, and I wanted to run it on ZOS Unix, so I had to change the connection string. But the program ran and it passed the report, put it into a DB2 table, and it was like five minutes of of work for me. Yeah. Uh so I I see it, it's very powerful for sort of prototyping or one-offs of these kind of things.
SPEAKER_03Yeah, I guess you need the attitude though, don't you? It's not like this can do my job for me. No. It's just got to be like the basis or to help generate the ideas and the initial stepping stones, and then you've got to use your your actual human elements to bring it all together.
SPEAKER_00Right, yeah. I use it a lot for prototyping. Uh, so when we're looking at sort of innovation items and we want to see how this actually looks like or would it work, etc. It's you can use it sort of to quickly build a prototype, and then somebody in development will have to create something that actually works.
SPEAKER_02I think that's yeah.
SPEAKER_00And and test it and so on.
SPEAKER_03Oh, yeah, testing. Do people still do that?
SPEAKER_00Check in production, test it. Doc documentation and so forth.
SPEAKER_02Yeah. Yeah.
SPEAKER_01So we uh We left you in technical sales, in pre-sales in IBM. So then what happened? How did you get into Rocket and what do you do now?
SPEAKER_00Um yeah, so so then uh I was there six or seven years uh and then in 2007 I I applied for a job at Rocket. Sort of just before that, uh Rocket was taking over the development, a lot a lot of the um the IBM brand, the DB2 tools. I ca I really like sort of the interaction with development, not necessarily being a developer myself, but sort of the interaction, sort of advocating on behalf of the customer, being close to development. And so I kind of wanted to be closer to development, so I I applied for a job in I don't know, late 2016, and I joined Rocket in in 2017, still kind of as a pre-sales um specialist, and then that's kind of what I am still today, although my title has has changed since then. So we help IBM at the back end with uh with pre-sales activities, post-sales, deployments, and and obviously being close to development, we we help with uh innovation and some of the new features and um being like the the customer advocate um to make sure that what we actually develop is something that the customers want and uh it's what they need. Um you know in in these days with with sort of more agile development and and uh the continuous delivery, obviously now there's there's much shorter feedback cycle compared to you know years back where you would have a new release of a product every three years, and and there will it will be much slower feedback cycle than than today. So there's a lot of interaction with customers um on new features and and what you know if we're doing the right thing and so on, including that modernization we talked about um earlier. So yeah, so that's that's what I've been doing since then. So I ended up being 16 years, I think, in in IBM and now I've been eight, nine years in in Rocket. I think I'll have 25 years of mainframe anniversary this year, I guess. Yeah. Yeah, thank you.
SPEAKER_03Doesn't make you feel old, does it?
SPEAKER_00Um well, yeah, I guess. I don't know if it's a positive or negative thing, but in in Denmark, uh you know, they raise the retirement age with sort of the based on your uh uh expected lifespan. So I can I can retire in at 71, I think, maybe.
SPEAKER_0371.
SPEAKER_00Oh my goodness.
SPEAKER_03And they might put it up before you get there.
SPEAKER_00So I have 20 years to go.
SPEAKER_01So you've worked with gosh, you said 25 years working with the mainframe. So what do you think have been, in your opinion, the uh the big developments, the the wows of db2 and the mainframe in in those 25 years?
SPEAKER_00I I think it's very difficult when you when you've been in the world, you kind of see sort of the small increments, and it's you you have really have to think about what was the major developments. Uh I as I mentioned, I like to look at new features, so that's kind of always where my focus is on sort of the the past new features. I I don't know with the actual mainframe, I think there's some major developments, it's kind of all incremental to me. But if if I look at some of the recent features, um I really like. And because I've never been a production DBA, I've I've only been an application DBA, I I never really had the responsible responsibility for a production system. I did break it as an application developer, that's a different story. Um but I sort of sort of some of the recent features I find really interesting is is the redirected uh recovery um uh feature of DB2. Uh in the past it was really impossible for customers to actually test uh recoveries now, unless they had like a DR environment where they could run these kind of things. Uh but now with the redirected recovery, you can run a recover test uh in your production environment and see, you know, demonstrate that you can actually recover. You are recoverable, figure out how long time it takes, um, and then sort of extrapolate to you know what is my um my RTO for an entire system. Um and we've seen quite many customers uh doing redirected recoveries uh and learning all kinds of interesting lessons about recovery. So I think that's I think it's a good thing. You know, we've seen customers that find out that that tape is slow. You know, if you have archive, it takes a long time to recover. Um and so the the um disk storage seller was happy after that. Yeah, of course. Um and I also think our customers found out that the R2 might be slower than they you know promised in some of their uh documents.
SPEAKER_03Or they were told. I think often it's just they were told what the RTO would be, and it's like we can recover in four hours.
SPEAKER_00Yeah, yeah, sure.
SPEAKER_03We have no idea.
SPEAKER_00Yeah, yeah, yeah. Now they have an idea. And also I uh some customers have learned sort of important operational experience with recovery because they have never done them before. You know, how do you restart a recover job? You know, what do you do when something breaks? What can break, etc. So I think for a lot of customers that has been a good way to learn without you know being in an actual recover situation.
SPEAKER_03Yeah, I agree that redirected recovery is a big step forward, but for me personally, I feel like it needs to go further because some places aren't going to be happy about you running a recover job in a production environment unless it's actually an emergency and you need to do it. Because it's it there is a risk to it still. If you get one little thing wrong, it could be wrong. Very simple.
SPEAKER_00Yeah, I I did have one customer that gets some got something wrong there, so that's one of the sort of the fall falls under the operational experience.
SPEAKER_03Right.
SPEAKER_00But I I I've I've seen some customers that are that are doing it quite well and they will set up RackF rules, etc., that prevents you from from sort of harming the production system. So there are there are some good ideas out there to make sure that you don't shoot yourself in the foot, uh, as I've seen some customers do. But I I do have some customers that have recovered 10, 20, 25% of production system over like a weekend to get really, really good estimate of of how how long time it will take to to recover their production environment in case it it uh you know they need to.
SPEAKER_03So yeah, that I think that's an exercise, isn't it? I mean that's a really good undertaking because you'd have to put a load more DASD there for that. And well goodness, yeah. Just thinking about all the all the operational and practical things that you need to do in order to actually achieve that. That's that's quite a major exercise.
SPEAKER_00Yeah, yeah. That particular customer, they had they had sort of capacity, at least CPU-wise, in in over the weekend, uh, so they had they they could run it there. But yes, they needed like 25% additional DASD. Um but that there's some good lessons to be learned uh of doing maybe you don't need to do 25%, maybe you can do five or ten percent, but there has to be some kind of non-trivial amount of data you recover to get, otherwise it'd be difficult sort of to extrapolate to to the full full amount of data.
SPEAKER_03Uh sure, yeah.
SPEAKER_00Yeah, but they're interesting. I I like I like that um feature a lot, and I I think uh those customers that have tried it out uh are also big fans of it. Um then another rather recent feature is the utility history. I know BMC has had that for for some years, but now it's in the IP. Nice to see you catch up. Well done. Yeah, I I think Hawkon Rupert has uh admitted that that has been an RFE or whatever it was called before RFEs for for quite many years. Uh but but it finally arrived in in DV2 version 13. And there's a lot of valuable data in that, you you you would know that. Uh but yeah, looking at you know uh CPU, elapsed time, etc.
SPEAKER_03Have you seen customers actually using the data in uh in a positive way to provide insights into what they're doing with their utilities?
SPEAKER_00Yes, and I think there have been some IDoc presentation, I believe, um, from some customers on that. Um I think from Commerce Bank. Uh I I I think I don't know if that's positive or negative, but I think some customers realized how much CPU copy is using.
SPEAKER_02Oh.
SPEAKER_00And so I think Hokon Robots commented on that, and they actually had to add SIP SIP uh enablement in the copy utility.
SPEAKER_03Win win then.
SPEAKER_02Yeah.
SPEAKER_00Because now it was very visible how much CPU you were burning on image copies.
SPEAKER_03Yeah.
SPEAKER_00But I I think there's a lot of valuable, valuable data in the utility history uh in in terms of also if you're planning to run a utility, you know, how long time did it take last time, and you can maybe extrapolate or do more clever things with with the data. So I think that's I I think that's a very powerful feature. It is still quite new, and a lot of customers had only made it to version 13. And it is something you have to enable, obviously, through a CPAM. So but I you know we hope that once customers start collecting this data, um, we have a lot of plans sort of to to exploit that data for for various things. But uh and the nice thing is that it's you know it's collected by the customer themselves, so it's it's very environment-specific uh for them. So it's very useful data in these kind of contexts.
SPEAKER_02Yeah.
SPEAKER_00Then maybe that's an older feature, but the Dv2 system profiles. I I can't remember when they were came around. They are some versions old. I don't know if you can remember them.
SPEAKER_03Mark's we're googling it.
SPEAKER_00Yeah.
SPEAKER_03Version 10?
SPEAKER_00Yeah, I was thinking version 10, but it might be version 9. I I think it's a wildly underused feature uh and maybe underappreciated feature. Uh I I've gotten a lot of questions from customers that said, hey, we have this remote application connecting, and uh they had a loop and they consumed all our threads or connections, or they ran some ugly SQL, and I I have no idea who it is and where they came from, and how do I get rid of them? And you can do that with the system profiles, um, you know, penalty box or even blog applications. Um and I I I talked to several customers that basically uh penalty box every application they have. Oh, so anyone that connects, they okay, you can have five threads or ten connections, but not more than that. I I asked them, hey, hey, what if you show good behavior? Can you get out of the penalty box? Uh no.
unknownStrict.
SPEAKER_02That's very strict.
SPEAKER_00Yeah, but it it gives you a way sort of to proactively um protect your db2 subsystem. And I I think a lot of customers seeing the distributed workload increasing and increasing. And it's obviously the for some customers it's the wild, wild west, at least large customers. You know, they have thousands of developers, and and the DBAs have no idea who is connecting to the system, there's no control, and the system profiles give them that control and and the ability to proactively manage it. You can also set special registers, so you can set Apple Compad for applications that may not tolerate a high Apple Compad, or vice versa. Uh, if you have the uh IBM and analytics accelerator, you can you can set the special register related to that. And there's all kinds of uh interesting things you can do on the system side rather than asking you know a random developer you you have no idea who is uh to set the right parameters um at their end. So I think that's an underused feature uh and underappreciated. I I I really like that, especially when I hear some of those customer stories, both from those that have done it up front, so they kind of save the DV2 subsystem, but also when those customers I talked to that have not used it, then understand, hey, you know, I now I could actually I could uh put that user in a penalty box or even prevent them um from connecting. So um yeah, I think that's that's a very nice feature.
SPEAKER_03Yeah, it's a really good point. And I don't think anyone's ever mentioned that one before when we've asked them. But I think you're right, it's underused and underappreciated perhaps and could be very powerful.
SPEAKER_00Yeah. And sort of in that DB2 has added a lot of instrument instrumentation for distributed workload. Uh so you get more and more insight into what is actually connecting from what IP addresses and users and so forth. So now you kind of have you have the instrumentation and now you have the sort of the hammer. So you know, so you can kind of create some kind of monitoring and and then also uh implement uh the profiles um and and and manage your workload that way. I think that's that will become increasingly pop popular as more and more distributed workload are are coming in. It's probably there that the the highest uh sort of the the growth in workload is from distributed, I think. Yeah. So so the the last thing I thought about was the very recent feature, and so and the jury is still out whether that will be popular or not. Um that is uh open telemetry tracing.
SPEAKER_01Uh right, that's quite new, isn't it?
SPEAKER_00Yeah. So IBM recently shipped uh open telemetry tracing for all of the subsystems, so db2, IMS, uh KIX, Kix Transaction Gateway, uh MQ, ZOS Connect, maybe something I've forgotten. Um so it's really sort of the the sort of the holy grail of the end-to-end monitoring that everybody has has wanted for many many years, but it was very difficult to follow a transaction that originated like on a mobile phone and all the way through whatever systems, all the way down to db2, and then kind of understanding what happened and if something is wrong, who who is the culprit? Um so it it looks very promising. Uh it is a new feature. Uh customer adoption, I think, is still unclear. There are some customers that are interested in it, obviously, otherwise IBM would not do it. Um but it yeah, the jury is still out on that. I I did together with a colleague uh Christoph, we we wrote a blog on on like how do you instrument a Java program and all the way down to DB2, and you can s you can follow follow that trace and uh and see everything. And uh Christoph, my colleague, is going to submit that for for uh Porto this year.
SPEAKER_03So oh that would be interesting. Yeah, it'd be good to see that.
SPEAKER_00Yeah. So that's that's a very recent feature. Uh so we we will see you know how how that is uh received by by customers. Uh it is very interesting. Yeah. Um yeah, and obviously AI is is uh is interesting as well. Um IBM has shipped an an agent, DB2 agent that you can connect to to your LM. So that I've I I foresee there'll be a lot of lot of things going on in that space. It's a very hot topic across everything AI.
SPEAKER_03Absolutely, yeah.
SPEAKER_01So, how do you think AI will affect the role of the DBA and the sysprog going forward? What kind of changes, what kind of DBA would use AI tools?
SPEAKER_00Yeah, that's a good question. Um so so personally I'm using AI a lot. Uh you know, I have Copilot home all the time, asking all kinds of questions. Um I I think like for for I mentioned Python, it's very good at write writing Python code. I I can also get some some OKJCL out for some some things. Uh but I I think mainframe is kind of um a niche area. So so I think so. I think those AI engines need to be, you know, you cannot use sort of the general ones to get good answers there. You need something on top of that, so like a mainframe-specific LLM that has some additional information. And I I think I know that there are many vendors working on that. So already shipped maybe the first version. So that that's going to be interesting to see. I I think it can be useful for any kind of person, even experienced people, because like just like things like looking up error messages, uh, you know, SQL codes, for instance, in the DB2 world, is today I go googling and I Google a what's uh you know minus six oh three or whatever, and then what is it and how do I solve it? So I I think everybody's going to use it. Uh that's sort of the information retrieval aspect of it, like hey, show get me some information, some generic information, but then you get into things like um, you know, how is my db2 doing? You know, do I have any problems? Um, these kind of things. Then you get into sort of more the AI agent world where you get information that is specific for your DB2 subsystem, and then you can combine that with the the general information, and then sort of the final step there is actually taking actions. You might you might ask the AI, you know, do any tables need a reorg? And it says, Oh, these five table spaces need a reorg, and then it's it asks us, should I run a reorg? Yeah, yes, sure, run a reorg. Yeah. It might be a while before we get there, but uh the technology is certainly there. So the the AI agents can can do these things in other areas. But I think in general the mainframe community may be more conservative, uh, so they they uh might be a bit behind on that. Uh but I I think eventually it it it will get there. And obviously a lot of us we have like a JCL library with all kinds of JCL for all kinds of things, but yeah. It might just be easier if you can, yeah, run the reorg, and I don't have to find my my uh JCL, not the reorg and uh and set the right options and so on. So I I do think it has great potential, but it's it's early days right now.
SPEAKER_03Yeah, and it would I think now or soon in the future it would be fine in the non-production environments. It's just taking it to the production environment and and the potential risks or just the fear of doing that kind of thing. Because probably a regular reel can run alongside stuff at a quietish time and it will probably be fine.
SPEAKER_02But uh I think that's where you kind of combine all of these things.
SPEAKER_00Um I I talked about the utility history, and so so maybe you it says, Oh, you know, you you need to run Reorg on these five tables, and then it could use the utility history and say, you know, last time it took two hours, uh, so you just you know you know how long time it will run. Then you then it would ask you and say uh build the the the utility statement and say this is what I'm planning to run. Do you think that's you know so I think there'll be a lot of confirmation steps and there's a lot of information we can augment into such a flow to make the the user sort of trust the process and and still be in in control. Um it'll be a while before we see like agents go off on their own and and then figure out whatnot reorgs need to be done and then go run them, etc.
SPEAKER_03Yeah, because it's also gonna need knowledge about the personal accounts application needs to be online between these hours. So you can't run anything between those hours. And you've got a maintenance window at this point, and if I start now, then the reorg is likely to finish in that maintenance window. So do you want me to run it now and it will be okay?
unknownYeah.
SPEAKER_03And yeah, I guess it needs it needs that knowledge as well, doesn't it? Not just a DB2 and not just about the the stats from what's happened in the past, but it needs that insight about the applications and the company and what's acceptable and what's not.
SPEAKER_00Right. And and that that's possible. Yeah. Well, and I mean if that information is available in that in that company and it's available to those agents, it can go find it. Like is if it's on a wiki or internal website, this kind of information. It so that that's it's uh I I think in other domains you can kind of see what AI can do. Uh and so I I think we just have to be uh innovative and then uh um yeah.
SPEAKER_01But yeah, though it'll the mainframe community is more conservative, so it it'll be a slower than on other uh other domains, but I think we will eventually the mainframe community is more conservative than others, and within the conservative mainframe community, I think DBAs and sysprogs are even more conservative. So I remember not long after I started working on DB2, which was DB2 2.3, so 3.1, 4, DB2 version 4, uh, we had to let users into the you know behind the gates and actually use spoofy for the first time. We were like, what would happen if we let users run actually run SQL against our database and not you know using predefined programs or or writing applications to get the data, which we could, as DBAs, we could analyze the uh the access paths. I mean, they could run anything. Yeah. We're sort of used to now people actually using spoofy and QMF and dynamic S. SQL. So it'll it'll come a point where we'd we'll say, w what would happen if you allow AI agents to actually access DB2 tables? They could run SQL and everything.
SPEAKER_00Well I I have at least one customer that is writing an agent to run SQL for an application area in production. They told me about the project. It was it's really, really cool. If if they get that on Rails, I I I'm I want them to submit that for iDoc. I don't know if it'll be this year or the year after. But they they they have a really cool um sort of uh process for they they they have the same concerns but they've kind of add the QA part of that process. So uh LLMs are quite good at generating SQL uh if you give them the DDL right so you give them the DDL they can generate a query quite easily um but then you don't know if it's a good query you know does it have all the right uh conditions in the join etc but then that then they're tooling to to do that QA part etc so they have like a quite nice pipeline um with a lot of uh sort of steps to to check the quality of the SQL and then um and then run it eventually um just through a chat interface uh but I that would look very cool to me so uh so I I think it's closer than we actually think um yeah maybe maybe yeah yeah I'll look forward to hopefully seeing that sometime soon or hearing about it at least so moving on a little bit you live in the Kingdom of Denmark what does a Dv2 geek do for his his time off when he's not working does he get any time off um yeah yeah I I do get time off yeah for many years uh I did uh triathlons oh then I I crashed on my bike and broke a hip that that kind of uh ended my running career um and then COVID ended my swimming career oh no uh uh and then so so now I'm and now I'm sort of just cycling uh so crashing the bike didn't end your cycling career no interestingly enough it did not end my cycling career I I was back on the bike I don't know half a year later that that was that was fine that must have been very painful though broken hip ouch uh well yes it was was painful um but I mean it did yeah well bad things happened uh it was kind of my own fault so that there's nobody else to to blame um and I it actually happened in in France my wife was not too happy because she had to drive home from France with our caravan and I I I flew home and and and she has forbidden me to go back to the French Alps um so since then I've only gone exclusively gone to the Italian Alps so for cycling because it's a massive difference. Yeah yeah totally difference wow oh that's that's quite a cool hobby though proper big mountain cycling I I do combine it with uh with with travel so I'll sometimes bring my bike when it when I can or rent a bike locally so uh for instance uh last October I was at the the IBM Tech Exchange conference the big conference in Orlando so I I rented a bike and and went out for uh a couple of days biking in the area attended some share conferences or been biking in Kansas and yeah all kinds of other interesting places it's it's a nice way to get out to see something else than just sort of the center of a city where the conference conferences that's very cool and especially when you travel somewhere in the winter like for instance Orlando it's it's the weather is significantly better than in Denmark in the winter time. Yeah yeah we feel your pain yeah so that's what I what I spend most of my spare time on right well so you keep pretty fit then that's good yeah fairly fairly I would say uh I do try to run occasionally running is so much easier when you're on business travel because it's easy just to bring a pair of running shoes um but my my running is is extremely slow some might call it walk probably still faster than us so likely yeah so this is uh this has been great to have you on uh on our podcast and we look forward to meeting you later on in the year in Porto yeah for sure I thank you before yeah maybe before P Doug or GSC Nordic or Escueladria where so I think Esquilla Adra is the same week as GSC Nordic and I yeah believe I'm going to GSC Nordic so I'm going to Helsinki instead of Croatia right it'll be beautiful it'll be lovely might not be as warm but it'll be lovely.
SPEAKER_03Right uh yeah I'm I might be at the at the police DB2 user group I think it's Gdansk in in June if I remember right yeah yeah I think I have something three weeks in a row in June uh conferences there so it's gonna be a busy month for all of us I think yeah well we'll look forward to seeing you um at those conferences and thank you so much for for everything that you do for DB2 and for the iDug community and for the local user groups and thank you very much for sharing your time with us this morning and and some of your your stories and your history it's been very very interesting yeah yeah thank you and thank you for for inviting me. You're welcome visit iDug.org for the latest DB2 community news, information about events and all of our technical content. Tune in again soon for the next podcast at idog.org forward slash podcast