85. Simplification of IT is Really an Illusion – Tech Field Day Podcast

Can there ever be enough simplicity in it? Once we've simplified things, why do we keep making them more complex? Join us to find out whether the dumpster fire is actually going to become the new normal, whether generative AI is going to remove all of our cognitive abilities and is going to do all of the jobs for us.

In this episode of the Tech Field Day podcast, Welcome to the Tech Field Day podcast, where we bring together a group of technical experts to discuss a single topic around some key issue in the IT infrastructure world. These podcasts are often recorded in association with one of our events. Tech Field Day is a part of the ING group, and you can also find this podcast on our sister site Techstrong tv.

In this particular episode, recorded on site at Cloud Field day 24, we wanted to take a look and think about the fact that simplification in it is really an illusion. But before we get into the conversation, let's meet who's on the panel today. Hi, I'm Kimberly Bates.

I'm formally of the Futureum Group a while ago, and I'm now an independent analyst, actually independent analyst hanging out in Colorado. Here we are. Hi, Nathan Nielsen.

I'm a principal solutions architect at Worldwide Technology, And I'm Guy Courier, currently a research director and analyst at the Futurum Group. And of course, I'm Alistair Cook, an event lead here at Tick Field Day, as well as doing the occasional moonlighting as an analyst with the Futurum group throughout Cloud Field Day. We saw a lot of conversations around simplification of things, making it easier to consume all of this complex infrastructure that we've built up over the last 30 years, trying to make it look like a, a simple service that's easy to consume, but is it really an illusion that this stuff's getting easier?

I think it's, uh, getting easier and harder at the same time. Cloud as a topic, that's a great example. Self-service.

Uh, you don't have to, uh, rack stack cable, no ticketing systems or fewer or whatever. That's the simplification side of it. Um, putting the power in the hands of the developer or the application manager to deploy infrastructure compliant infrastructure, whatever it might be, that's all simplification.

But then that easy on-ramp suddenly allowed there to be 12 different services per provider across four providers, and then on-prem still needed to be necessary. And so the multiplication of possible choices and the need to provide, you know, consistent capability services and compliance across all of them. Um, it's almost like an arms race of its own, in a sense, in a sense.

And I actually think the increasing choice, the increasing options, the increasing complexity outpaces the simplification efforts that we see. I'm, I'm seeing a, a similar source. So my, my primary area focuses more on like the, the campus, the on-prem networking portion.

Obviously there are, there are applications that are being utilized that are not on-prem, but the, just the, the complexity of the network itself is, is growing significantly protocols on top of protocols, underlays and overlays. Uh, but all of our, like the major OEMs are driving towards a unified, and I'm not gonna say the term because, but because there isn't a single one, but it's, it's a centralized management platform that, um, that does simplify the implementation and the management and things like that. There's a lot more going on under the hood, but the idea is that, okay, there's gonna be a little bit more point in clicking versus having to look at 500 different moving parts and trying to figure out how to keep this, this train going down the tracks.

So I think I got the short stick on this one because I, we, we were supposed to be debating here, and I, I will, I would absolutely, would completely agree with both of you guys, but then I, I will take the, um, the other side side of the argument is like, do you guys wanna go back to assembler in the mainframe? Yes. Yes.

Those are fun. That's fun. Thank you for asking.

Hey, hey, we have web assembly now Gladman one, one of my favorite things too, I mean, when we're really thinking about, it's, if we go back to the complexities and the architectures of where we come and, and you know, to be fair, what we have done is we have advanced so far with what we can do, and every time we take on another level of something that we can do, which is okay, so we got so simplified that we took whatever was in the mainframe kind of thing, and we stuck it in this thing called the pc, which at first was a suitcase, and then it became whatever it is today. I mean, you think about this, how that has so simplified what we do from a word processor to, to word to from calculators and your HP little calculator that you have to, to excel. I mean, it's amazing what we've done in terms of taking very, very complex things and simplifying it.

The problem is, is that we keep moving the goalpost, And now the goalpost is this thing called, well, we added the cloud. Okay? So, but then we decided that we can't just put it all on print.

We can't just put it all in the cloud because there's all these other reasons. So then we gotta go figure out how to make the two work together. And so we made it complex and, and let's, at the same time, let's throw an AI there.

Oh my gosh. So part of it is moving the goalpost. I mean, I know about you, but you know, when I had a teacher that moved the goalpost, it p****d me off.

I don't know, I, I, Kimberly you can measure complexity and I'm sure there are academic studies and mathematicians and so forth out there with above my pay grade with above mine, trust me. But with quantitative measurements of complexity. And if you just, I I, I will say that I think you can define complexity in terms of the number of parts that need to interact with each other and how many dimensions on which, or how many, how many different ways they need to interact.

Mm-hmm. And, uh, my initial premise was that, um, there are new ways to, uh, unify access points. Like the example of an application manager deploying an environment for a, um, to, to host, you know, an application, a database, whatever.

Uh, uh, there's, there are fewer access points to do that. There is a, an interface to allow you to do a lot of things, whereas there used to be four interfaces plus a ticket that is, you know, one point where there used to be six. Mm-hmm.

The problem is that underneath that or above that, there are 20 options where there used to be one. And so those are those, that's that addition of points. That's the multiplication.

I think you can measure this sort of thing, not that we're doing that, but if you could, I would say the complexity is increasing. So Talk about your network world. So yeah, that's, I mean, he's spot on from a networking perspective as well.

Like, I, I think that it's, there, there's a correlation between capability and complexity. Like as the capabilities increase substantially, the, the complexity to maintain the capabilities is also right there along with it. Um, so it, but at the same time, the, the need for simplicity is also like front and center.

So it's, hey, we've got, we've got this thing doing 15,000 stuff, but 15,000 things. But, um, we, we need to make sure that we are only manipulating it from this perspective because so many other things need to like, fall right in place in order for it to, to actually function correctly. And there's a whole cognitive element to this, is that humans doing things, we can only cope with a certain amount of information, a certain number of work items.

You know, we have limited working memory. And, and so we've gotta hide away a lot of those dimensions of interaction that guy was talking about in order to be able to cope with handling this. And this again, is, once we do that, once we start making it easy enough to manage what we had before, we just layer on doing more.

We, we always have an, an objective to, to do more. Uh, I thought it was really telling yesterday when we were with Brian Rell looking at his collection of very ancient documentation of early systems, and that was where it was assembler and mainframes, uh, conversation around assembler and mini systems. You know, we've come a very long way.

There's a lot of things that we no longer have to spend a lot of time thinking about, but there's a whole lot of new things that we have to think about. Uh, some of that's the, the transition from new ideas get leading edge and, and, and what we're seeing with AI at the moment, lots of leading edge, but bleeding because they're not, you're not building in things like, well, security and governance because you're so fast building the new thing. Well, eventually we'll build in that security and governance and it'll just become normal service.

So a friend of mine, Frederick Van Herron once said that AI is simply something that we don't understand how it works. Mm-hmm. And um, and the example that he gave is that speech recognition at, you know, when you looked at speech recognition many years ago, we call that was ai.

And today it's just speech recognition. It's, you know, how well does it actually interpret it? Um, so, and that to me is an example of how the goalposts have moved.

We simplify it, then we invent new things and we simplify that. Um, the question becomes is can I, I think that I have on this is can we continue to layer on layer on layer or do we need to have at times, like oxide is talking about a clean slate in a sense the cloud became a clean slate or on-prem, so, and you say, oh my goodness, this is really simple. But you know, if you went to AWS reinvent over the years, you got overwhelmed by all the announcements they had.

Yeah. And then that one year that I went and I listened to them talk about the well architected environment, and they put the charts up there and I went, holy crud. That's Complex.

Yeah. Yeah. So yeah.

I mean, I, I started learning about AWS when there were 46 services. Yeah. Um, you can tell I'm, I'm old now and, and there's more like 300 services.

I lost count of the number. And yeah, it is that thing. We started with something simple and then we kept finding use cases where we needed just a little bit of a change from that, just a little bit different.

And all of a sudden there's yeah, 300 services and there are 400 different types of EC2 instance that you can buy. It's no longer t-shirts. Yeah.

So that's an example of the quantitative thinking around is it more complex? But I think one part I missed is, you know, 400, you know, shades of something otherwise similar would be simpler in some ways than 400 wildly different things. Mm-hmm.

So it's the, it's the differences among them. Um, plus, and, and you know, I'm so glad that, that, that, uh, you're hearing Nate, because the networking perspective is so important and, and I, I know that, but I so, so often overlook it because the integration interaction of these disparate things Yeah. Especially across the life cycle of an application.

Um, and I, I focus, there was a nice focus during cloud field day on the application and the, the, the, the outcome. Um, I focus on that because, uh, we, we, we, we tend to think about how to twiddle the dials better to bring back a, an old phrase we've used on this podcast. Um, we think about that, but ultimately the outcome, um, is for, uh, to take advantage of all these new capabilities while maintaining a reliable, relatively straightforward and high value output, which is the functioning of the application and the use of that application by the rest of the organization and the people and the partners involved.

And you know, how that's delivered, how that is, how those integrations occur, um, is critical both to that functioning and to the level of complexity that we're dealing with. Yeah. You know, to, what that brings to mind for me is, is like the, uh, the, It's so hard to contemplate on getting, contemplate contemplating it.

Go ahead. The, the, the rapid advancement of, of AI ops, and within the networking industry, the whole point, it, it used to be like, Hey, machine learning and, and AI is definitely embedded in, in these things. And that's just to help us correlate data so we don't have to go track down logs and, and net flow and try to like piece all this stuff together to come up with, with something.

We've got, we've got a machine that's helping us do that really fast. Now it's moved to the point where it's doing predictive analytics where it's, it's taking dynamic baselines and it's looking at things that are perhaps an anomaly based on what it normally is. And because of this anomaly, there's a good chance that this thing might happen.

So you wanna go take a look at that, and they're helping you get in front of problems before they're actually problems. So there, but there's a lot of, there is a lot of complexity in that, but it is at the end, simplifying operations, speeding up mean, the, the meantime to innocence, meantime to resolution. Um, and, and it's, and yeah, there's, there's even, what, what is crazy to think of, like self-driving cars, self-driving networks are the next thing on the horizon.

There's already, manufacturers are taking steps towards getting to that. I mean, right now it's just, Hey, I noticed every time this, this, this thing happens, uh, this device stops working. You have to, you have to shut and no shut the port to kind of just bring it back to life.

Next time it stops sending data. Would you like me to just disable and re-enable the port for you? And you'd say, yeah, that's a good idea.

Now the network is making a suggestion to you through natural language and you're typing, yeah, that sounds like a great idea. And now it's scheduled a job to watch for data to stop going on this link, and it's gonna shut, no, shut the port for you. So it's, it, it's wildly complex stuff that is simplifying day two and operations and, and allowing us to be, you know, 10 times more efficient with a lot more complexity.

Yeah. So sometimes refer to it advanced simplicity. Yeah.

Mm-hmm. So very, so that is incredibly complex underneath, but that is simple to operate. And I think that's, that's the sort of spaces that we get, keep seeing advancement.

So, And I would think from A-C-I-O-C-T-O perspective that's looking at their operations is to look at what is on the horizon that we can potentially be getting ourselves into. Okay, so let's say we get to a thousand points of networking, right? Or, or even more than that, you know, thousands and thousands and thousands of points in networking.

The CIO and CTOs need to be thinking about where do we need to go to be able to support this? How do we strategically get out to that point and manage those environments? Um, and I think that companies, the vendors that are thinking that way, um, versus trying to patch the problem that we have, and hopefully they are, um, it will be the, the really super successful ones that are going out there.

Um, I, I, I think so, but I think they need to do both. There's a lot of need for the patch. Mm-hmm.

And, um, actually it's really interesting that you bring up the, the, the C-C-I-O-C-T-O perspectives and the ciso, because what Nate described, I mean, AIOps does, we are tempted to, to say that, that the, the implementation of AI into operations of all kinds simplifies it. Then we forget at that moment, and we remember later, especially when it bites us in the butt that the implementation of that AI into the system engenders complexity. We saw a great example from Fortinet mm-hmm.

Of how, you know, AI creates security risks that we can't even imagine that in fact, the hacker doesn't even imagine necessarily the hacker ask the ai, Hey, how can I hack this system? And the AI tells the hacker. So that's, maybe that's the goalpost moving that you talked about earlier, isn't it?

Mm-hmm. That's the goalpost moving. Well, that's part of the goalpost moving the, the other goalpost is the people that come up, you know, open AI is a goalpost.

Right. Those were goal, I mean, open chat GPT, we woke up in November, whatever year that was, we went 2023. Mm-hmm.

Holy Cow. We'll Never forget. Yeah.

And now we need to re completely rethink. I mean, at that point in time, the CEOs didn't even know what was going on. And, you know, three months later they're getting asked by their board, what's your AI strategy?

Or whatever's going on. And that's, that's a goalpost mover, right. Um, and with that comes the responsibility of, you know, the technologists for, you know, how how do we keep, keep it simplified as we absorb this technology?

Mm. Yeah. Especially when there's so many unknowns.

Like it's, it's not that it's just something so new, it's, it, there's literally built in intelligence within, within IT that we're just still getting figured out. And Yeah. It's, and in fact, even, even the networks themselves have had to change just to support it.

So, so they, there's a term that we often refer to. There's both AI for networking and then there's networking for ai because it's, it's just a completely different thing. It's, it's not the same networking that has been the same for 30 years.

It's a completely different thing. Yeah. And that in the simplification, one of the things, especially, you know, as I listened to you talk about, you know, kind of the work you do, having AI ops in there to do some of that work for you, frees up when we talk about frees up time to be more efficient.

Well, I think it also frees up time to think, you know, and you and I talked about needing that time to think, to sit down and do the research we do, and 15 minutes is not enough. You know, I need several hours to go and dig deeper and be that, do that kind of work. And I think that's super important for, you know, any it individual that is working in a position of authority at the company to say, how do I think through where we're going and what we need to be doing, and not at, just to the point that we're trying to refresh the hardware or the software or something along the lines as part of their job is to think through that.

If I, if I may, I'd like to provide a counterpoint, because there is one thing that simplifies, and that's ignorance. Ignorance simplifies. And I'll give you an example.

I'll give you a concrete example of that, which is, which is telephone, phone service. Um, if you lived through, like I did the transition from Plano telephone service to cellular service, one of the, uh, outcomes of that was service degradation dropped calls. We live with that to this day.

There was a time when to not get a dial tone when you picked up the phone, at least, you know, in most of the world, was a remarkable occurrence that happened twice in your lifetime. Now, drop calls or service unavailability or any of that sort of thing is just taken for granted. And that is a form of ignorance, which is to say that, that, um, we all stepped into the advantages of this highly complex system that gives us connectivity wherever we are.

And when there are problems, everybody just knows it's just a problem. So what I'm trying to, the the analogy I'm making is that if you take ai, the running these highly complex systems on our behalf, that we don't have to, and we can spend our time thinking or doing other stuff instead, then when there's a problem and a blowup an entire region goes down, or an entire application or whatever, it's just like, oh yeah, well that's, that's just how it is. And we, and we don't think about the, those good old, uh, assembly mainframe days when it just didn't happen.

Come back tomorrow, later this afternoon, we'll be back up. Yeah, that's a, that's a simpler world, right? I don't have to deal with it anymore.

And they've always said, you know, the cloud means I don't have to care. And turns out that's not true. You are still gonna care though, But I'm no, who's gonna care?

Who's gonna care? I, I'm, I'm, I'm asking a serious question. Who's gonna care?

Yeah. Oh, hey, I just lost a million dollars for us today because of this AI system that I've been supervised. Yeah.

You commerce, people would care. Yeah. Your banks would care.

Not if we have insurance. I don't know. I'm just playing at playing this out.

So I think there's an analogy to what Nate was talking about, where the AI is doing a shut, no ish disconnect, unplug and replug the network cable for something that's gone faulty. And it's now actually just hiding the fact that the thing over there is faulty. Right?

Right. There You go. And so, returning to guy's point, does that matter?

Right? It, if you look at it in isolation, it's a bad thing because we are just patching, we're just adding more debt, more technical debt and patching the problem. But the outcome is we're still delivering the service that's required, and we've done that with a minimum effort.

So there's an element around, you know, what's, what is now acceptable? What delivers value to the business that it has the, the least cost to it. Um, I, I'm trying to get to a, to describing a dumpster fire for, uh, as normal for, um, for guy, but I haven't got the analogy right yet.

But that, that feeling of, if it's perfectly normal for the, the dumpster to be on fire, then it's not a problem that the dumpster's on fire, which is your point about if it's perfectly normal for the calls to be dropped, it's not a problem that the calls are being dropped when it used to be a big problem. In fact, I use it all the time. Oh, hi mom.

I'm sorry, I didn't hear what you just said. The, the, you know, the signal's not very strong, you know, so I think, sorry, mom a little Bit, you know, and, and you know, one flip, I can't, I guess I can't use that anymore. One, one flip side to that is that we, without doing the thinking, so, so there, there's two thoughts that come to mind.

Number one, it's, it's, we are changing our mindset to be in more of a proactive state of mind versus reactive. And like, oh, everything just went down. Let's figure out what happened.

It's, Hey, let's take these measures to not let everything go down. But, uh, and, and then the other thing too is that when it's not taking any mental effort to do this task, you're also not learning anything about that task. You don't, if when it does fail, you have no idea why, because you didn't know how it worked to begin with.

All you knew is that if I click this button, all that stuff is gonna happen. So yeah. There's also a, a, i, some people use the term that AI's actually making us dumber, and it's because we're not thinking for ourselves anymore.

Just like, you know, doing simple math on your hands. If you're, if you're not used to punching into a calculator, you're not calculating yourself. And yeah.

So, so there's, there's downsides to simplicity too, because it, it can lead to ignorance. Yeah. And, and a broke down in critical thinking, which is what's being seen in young people who are coming up through education and using generative AI to do the critical thinking for them or to emulate critical thinking, their own ability to do, do the critical thinking is, is degrading.

And hopefully that's not gonna be a, a long term trend. Hopefully that's something we're gonna be correcting on, because as Kimberly says, you need to step back, take some time, think about what's going on. But what usually happens at that point is you build the new complexity mm-hmm.

That takes up the time that you used to have to, to think about things because you had had simplified previously. Yeah. It's a still, it's a human condition thing where this is, this stuff doesn't happen in isolation.

It is all driven by us as human and our human behaviors. I, I think we might be in a moment now where, um, the, there, there a generational shift in it around the standards. We had a similar shift, I think, right around when client server came in.

Mm-hmm. Um, what standards are we gonna hold these systems to? And I think it's evident in the, the split between, let's say the more scientific community or the more academic or government community of it, which is, which remains very strict and, and, uh, holding very high standards and having a great deal of collaboration and a great deal of systems related work versus, you know, maybe it's the Silicon Valley ethos really coming out all over the world in all sorts of areas, which is the move fast and break things approach.

These both have existed for a really long time, but, um, I think there was a, there was a fair amount more rigor in, in the, the, the, the, you know, incoming gen, uh, sorry, in the, in the current generation of it. I, I wonder if we are starting to feel that, that relaxation around, you know, availability, reliability, uh, data protection, all of those other sort of things that I think we hold very dear network, um, performance and, and, and availability. Um, maybe, maybe in five years we're, we're gonna be looking at it very differently.

Well, I think we are running to our end of time. So, uh, this conversation, of course, like all conversations at Tech Field Day events could go on for hours and hours. And, uh, we probably will talk for hours about these topics following this, but if those, those of you watching the video, would like to, uh, engage some more carry on the conversations.

Kimberly, where can people find you? Oh, I'm on LinkedIn. Kimberly Bates, go up and find me there.

I'm, I'm on LinkedIn as well, Nathan Nielsen at wwt. com. I am also on LinkedIn fairly frequently, uh, LinkedIn slash in slash Guy Courier and hit me up on Blue Sky Guy Courier at Blue Sky Social.

And I'm Alistair Cook, an event lead here at Tick Field Day. You can find me also on LinkedIn and various socials. You can find all four of us on the Tech Field Day website.

Just look for the Cloud Field Day 24 event, and you'll find not only my three guests here, but the remainder of my delegate panel for the event. Thank you so much for joining us for this episode of the Tech Field Day podcast. Uh, you can find us both on YouTube as well as in your favorite podcast application.

Be sure to give us a nice review and, uh, and, uh, a like along the way. Uh, tech Field Day is of course, part of the Futurum Group. You can also find this podcast on Techstrong tv.

com/podcast. So thank you so much for joining us. You, we will see you again next week on the Tech Field Day podcast.