Introducing Telemetry News Now

Welcome to a very special episode of Telemetry Now. I'm your host Philip Gervasi, and joining me today are Leon Adato and Justin Ryburn to help me introduce to you a brand new podcast, Telemetry News Now.

Telemetry News Now is a biweekly tech news podcast coming this fall on opposite weeks of our main show, Telemetry Now. Leon and Justin will be cohosts along with me, and we'll be bringing to you some of the more important and interesting news items of the week with commentary, hopefully, some useful insight, and perhaps some witty banter as well. So with that said, and instead of long introductions, let's get started with today's episode.

Leon and Justin, thanks so much for joining me today on this very special episode of Telemetry Now. Neither of you gentlemen are strangers to the podcast, of course, having been on as guests at least once. Justin, I think you were on recently to talk about BGP flow spec. That was a great episode.

Today, though, we are introducing a brand new show, Telemetry News Now, hosted by yours truly and Leon and Justin as co hosts. So Telemetry News Now will start up this fall, as I said in my intro, and come out biweekly opposite our main show, Telemetry Now.

But before we get into any news today, I would like to spend a few minutes for you, the audience, to get to know Leon and Justin a little bit more. So Leon, why don't we start with you, a little bit about your background, who you are, how you came to be in tech, maybe how you got roped into this new podcast.

Okay. Let's see. A little bit about myself. I like long walks on the in the data center, warm aisle. I'm definitely a warm aisle person, not a cold aisle person, in the days. So I've been in tech for, like, thirty five years.

You know, for those people who need a more meaningful reference, that's when Windows two eighty six came for free on twelve five and a quarter inch floppies when you bought a copy of Excel one point o, which nobody did because Lotus 123 was the bomb, and, you know, everyone used it. The point I'm making is that I've seen some things. I've been around. And so I tend to really enjoy, news articles that are old.

They talk about old stuff because I've seen it, you know, come on and then sometimes go back out again or whatever. Those really intrigued me. I've been a news hound for a really long time. I mean, you know, back in the nineties, I would buy or get a subscription to, you know, PC Magazine and Network Computing and InfoWorld and, you know, and then I would be tearing pages out of those newspapers and sort of marking them up and sharing them with coworkers who usually didn't wanna know what I had to say.

Also, as you listen to more of these episodes, you'll know I'm an AI skeptic.

I really I'm not I'm not on the AI bandwagon. I'm just not I'd and and it's not that I think AI itself is bad. I just see it poorly used in so many cases that I just sort of roll my eyes and you can hear it.

And I'm also a remote work evangelist. I've been working remotely full time for over twelve years now. And before that, I was working at least half of the time from my house, so I tend to jump on anything that enables or talks about remote works. That's that's who I am. That's what I do. As far as why I love this kind of show, I I think it's really hard as an IT practitioner to keep up with the news. There's just so much of it, and it's you know, all the social media networks are fire hoses.

So I think that if we can do some of your pre chewing of your meal and providing you with the stuff that you're gonna care about, I I consider that to be a useful and friendly service.

Great. Thank you, Leon. Much appreciated. And I do look forward to hosting Telemetry News Now with you. Justin, a little bit about yourself.

Similar to Leon, I've been in tech about twenty five years back from, when there was no such thing as specialties in technology.

You know, people did everything from storage to compute to networking to security, application development. It was all basically one big discipline.

But I kinda fell in love with the networking side of things as specialties kinda became a thing, all the way back to the early days. And, you know, prior to graduation from high school, my buddies and I would dial into a BBS and ultimately get access to the Internet. And I just remember being fascinated by how quickly you could share information, how quickly you get access to information. I I remember thinking, like, this is gonna change our society. Like, this amount of quick and easy access to the Internet and to all this information is gonna be, like, you know, it's gonna completely change our society.

Little bit of a finance nerd as well. I when I was in, college and then again in grad school, any electives that I had could have, I took, finance electives. I've always liked markets and financial markets and investing. So, you know, I'll probably bring some news articles to the table that kinda combine business and finance and and tech altogether.

I love the intersection of business and tech. I I have chosen presales for the majority of my career mostly because I like to hear about what customers' business problems are and figure out a a technical solution that can speak to their to their problems and help try and solve those. So, you know, I always look for articles that, the articles that I find interesting or articles that combine those two things, business and technology.

And then, the last thing I'll say is I think, I take the opposite approach of Leon when it comes to AI. I think it's gonna be the next Internet. I think it's gonna revolutionize our society. It's gonna be the next big thing. So I think we'll have some interesting debates here on the show on, some of the stuff related to AI as we go forward.

Great. So why don't we get started with the news? Leon, why don't you get us, started off for the day?

So I mentioned that I like, you know, articles that reference old technology, and, there's an article out that WordStar seven, which was the last version that was run completely on DOS, not command prompt, get off of my lawn, a run on DOS is being rereleased for free.

So people who wanna relive the glory days of control KQ and all the other commands. This was this is really special in my heart, not just because it's old, but because WordStar was the first word processor I ever learned, you know, back when I was in college, actually. And so there's there. I I wouldn't willingly go back and use WordStar a lot.

I was more of a word perfect kind of person, but there there is a special place in my heart for it because it was the first word processor. And to hear that, you know, word star seven for DOS is being rereleased for free just made me chuckle. I also am certain that the only reason they're doing this is because George RR Martin hasn't put out the last, of the Game of Thrones books, and it's well known that he will only write in that word processor. That is the only word processor he uses.

So I'm sure that they are trying to encourage him subtly by saying, here, have the software.

Please finish our book for us along the way.

Yeah. That that really speaks to that whole nostalgia thing that we have in our industry. I don't know. Maybe maybe people in other industries have it like lawyers and plumbers and x-ray technicians.

I don't know. But we certainly do. Like, isn't it awesome like when you see a picture of like an old like Amiga computer and you're like, oh, look at that. I mean, I know for me, I use TeraTerm a lot for terminal access.

And, I I immediately changed the the colors and the background and, you know, the way everything looks. So it looks exactly like it did when I used it first, like, twenty years ago when I first got into tech. I'm not sure. I think it's just a thing of human nature where we look back with fondness on things from our youth.

And to be fair, with rose colored glasses. Right? Because it's not like this word processor program that you're talking about is better than anything else. I mean, it's probably terrible compared to what we're using today.

Right?

No. No one will ever try to tell you that control k q is a good way to exit a piece of software, except maybe people who use Emacs.

No. But, I mean, look at how many people look at how many people play retro video games. Right? Like, there's a huge population of people who love retro video games. There's a whole population of people who would argue to the death that, vinyl records are sound better than, you know, high quality digital audio. And some of it is you know, there is something to be said for that, that media, but then there's just something that's you know, people like what they grew up with. Right?

What they're what they're familiar with from their childhood.

I'll add one more thing, though, from a technolog technologist standpoint. I think it also has to do with the sense memory and the muscle memory that we build up. Oh, sure. We do especially as network engineers, right, we do really complex motions.

It's not just commands, but motions of commands, sequences of commands. And when you learn it when you learn it and you do it over and over again in the original context with the screen looking a particular way and the font looking a particular way, and then to try to do it in this different, you know, different colors, different font, different size, different layout, it doesn't your muscle memory is sort of jarred a little bit. So by putting things back the way that you had them, it makes it easier to do certain actions, which I find fascinating from a psychological standpoint. Yeah.

I wasn't joking when I said that I changed the font and terra terms so it looks in the colors too, so it looks the way I want, so I'm most comfortable. And I feel like I type faster, and I feel like I remember more commands, which is completely false, but I feel like I do. But you know what? To be fair, even when I get a new keyboard, right, like, my Logitech, like, falls apart or something, I had a new Logitech. It takes me, like, a week before I'm comfortable and things work properly.

I'm like, wait a minute.

How does this thing work? It feels different.

We could probably do a whole podcast on preferences for keyboards. Right? Whether you like the ones that they're clicking, you know, whatever. I don't know.

I I just use whatever's in front of me. I'm not I'm not into that, but I know people are like, you know, fight to the death. This is the keyboard. I I'm sure.

Well, WordStar, we salute you. Hanging on. Hanging on maybe by a thread.

Anyway, Justin, let's move on to our second new story of the day.

The EU and the UK have cleared the acquisition of Juniper Networks by HP. I know probably most people listen this podcast are aware by now that they the HP has entered into an agreement to acquire Juniper Networks. The acquisition like that, of course, takes a lot of different regulatory hurdles to get through.

Most folks were saying they figured that, this hurdle where the UK and the EU approving it was gonna be the biggest one to get it done. So, it appears that this is getting more and more likely that they will hit their their target date of having this deal closed in early twenty twenty five.

You know, this one I I picked because it does have a special place in my heart. I spent ten years at Juniper and, just love some of the innovation that they have done over the years. For those of you who aren't familiar, Rami Raheem, who's the CEO at Juniper, started off, as an intern and actually worked his way up. So it's kinda just kind of a great story of starting as an intern, working his way all the way up to the CEO. So I think it will be interesting as the two companies presuming this goes through, and the two companies get merged together, I think it'll be interesting to see what they do with their overlapping product lines.

Yeah. This is a this is a fourteen billion dollar acquisition. So this is not a this is not a trivial matter. This is a this is a very big deal for HP and for the tech world in general.

I mean, it it dub it pretty much doubles, HPE's operating income. So that's a significant change to what HPE is all about, and I'm sure that's gonna change the drivers for the company moving forward. It also, significantly, changes and boosts, HPE's networking business. So that's also gonna change the dynamic of the company, moving forward as well.

And right now yeah. Sure. It it is approved in the EU, and, and we just have we have to wait now for the FTC here in the United States. And, I don't know what the expected close date is, but from what last I heard, we're looking at the very end of this year, twenty twenty four, possibly, early twenty five twenty twenty five. So we we will see. I mean, another thing that's interesting to me is also how there is now more overlapping business between HPE and and Juniper.

Juniper traditionally in my mind has always been a service provider, networking, vendor, but that's really not the case. I mean, when you think about some of their SDN products, early on with Contrail and their missed acquisition and, and then moving into wireless and and the things that they're doing now, in security as well. They certainly aren't a service provider only networking vendor. So there is gonna be some overlap here between other acquisitions that HP has made over the years, both in the wired and the wireless space and more or less in the enterprise world as well. And, and then the product lines that are gonna be absorbed through this acquisition.

Mhmm.

Yeah. I mean, you know, both HP and Juniper have stated that they don't really know yet. Right? They'll figure that out.

They're gonna try and do what's right by their customers. I think as network, you know, practitioners, you know, listen to the podcast, that's a great question you wanna ask as this gets closer to closing of your account teams on both sides. Right? Depending on how long you plan on having gear in particular part of the network that you're doing a network design for, EOL of equipment can be something you wanna consider in your design.

Right? So I think that's something you wanna pay attention to, you wanna keep on your on your radar as a, you know, as a potential design constraint.

Do Do you think that this acquisition might be a move by HPE to expand beyond, enterprise networking? I mean, I remember installing or really, to be honest, replacing a lot of ProCurve switches in my day.

But this is an opportunity, though, you know, I did say and you agree that, Juniper is not exclusively service provider. They do have a big, market share of the service provider networking market. And so this could be a move by HPE to sort of, get a really strong position in both the enterprise and the service provider space.

Yeah. I mean, it you know, the they're they're getting a lot of assets that are definitely divergent from their historical, product line. Like you said, the service provider stuff, Juniper has custom silicon, which, HP has typically used merchant silicon. They bought it from, you know, Broadcom or whoever else. So, you know, there's a couple different assets that they'll get here with Juniper that they don't historically have in addition to the overlap in some of the enterprise product lines like the switches and the wireless. So, yeah, it'll be interesting to see how much of that, they continue to invest the r and d dollars in and how much of it they EOL or dive divest. It's gonna be interesting to watch for sure.

There's also the other, component here, which is HPE making a move to acquire a company that already has what, I guess, we can consider a more mature AI product, AI for networking product.

Now Juniper, through well, really, its own, R and D over the years and then especially in more recent years through the acquisition of Mist, which I I mentioned earlier, really has a very strong position in the incorporation of, artificial intelligence. And what I'm gonna say is more, you know, the application of ML models and statistical analysis and all of that kind of stuff to the world of networking. And we're seeing that now being applied in all of Juniper's product line, not necessarily all, but in a lot of Juniper's product lines. So they're taking that Mist technology and, and incorporating it, which is awesome. And so, not that HPE is just jumping on the AI bandwagon because there is certainly value to to, to incorporating this technology in the networking space in our industry.

And so, this is certainly a mechanism or a means for HPE to to also move in that direction through an acquisition and, and and not necessarily in terms of service provider or enterprise, but specifically in terms of expanding its AI, offering.

Yeah. But I I think your point is very valid that people have have this bias in their head that Juniper is more of a service provider business. But like you said, a lot of their acquisitions over the last five to ten years have been very focused on enterprise. Apstra is another one that you didn't mention. Right? It's very very focused on data enterprise data center type of environments, orchestrating the configurations, which is that environment stuff. So, yeah, it'll be interesting to watch, you know, what they do with all the pieces here.

And then there's also the human factor that, sometimes we miss, maybe. I mean, I think all three of us know folks that were and maybe still are at Apstra, Juniper Apstra now, and they went through that acquisition. And for some folks, you know, that might be, a great new opportunity. For some folks, they don't care for it because they go from a smaller company company to a much larger.

I believe Apstra was somewhere around, like, two hundred people when it was acquired, give or take, by Juniper, a significantly larger company. And then here, again, through, yet another acquisition. And so the question arises for folks, I assume, because I haven't gone through this. Where will my job be here in a year?

Where is my career going? Perhaps this is a great win for somebody, and perhaps it's a it's a time to make a change. Who knows? And then for the consumer, you know, we have to start thinking about, okay, what will be end of life, EOL, in the in the near future that I have to consider?

I I had this huge outlay of capital for hardware in my data center, in my campus for my firewalls, and, I need to, depreciate that over five years, seven years for it to make sense. Will these devices be supported for that long?

So in in this case, the size and scope of this acquisition in these two companies, to me, suggests that none of that stuff is gonna be an issue because you are talking about a huge amount of personnel and skilled staff that are gonna support product lines that are out in the field that do generate revenue for this new company, HPE, and then Juniper acquired by by HPE. So I don't I don't see that as an issue necessarily, though, of course, that's that is speaking of the near term as in, like, five years.

Yeah. And, I mean, I think to Leon's point, we, you know, we spend a lot of time reading and researching about the newest, latest, and greatest technology, but they also have to remember there's a long tail out there. Right? There's still a lot of people who are buying in a lot of bulk, stuff that may not necessarily be, you know, real exciting cutting edge stuff, but yet there's still a lot of money being spent on it. Right? So there's still plenty of money to be made selling, you know, printers and, switches and wireless and a lot of things that may not seem really super cutting edge to us, but there's a lot of money being spent on that.

Yep. Absolutely. There is a divergence between, what I think we see in tech news and then in the, the the thought leadership, content from all of the, you know, pundits and talking heads out there in the tech space, and then also what people are actually buying day in and day out, to run their businesses.

In any case, let's move on now to, the next, news item of the day.

So a recent Gartner report predicts that thirty percent of generative AI projects will be abandoned after proof of concept by the end of twenty twenty five, the end of next year. We are here, nearing the, nearing the third quarter of twenty twenty four.

And, we're hearing that the large players, the big web scale companies that are doing a lot of work in, generative AI in particular, Amazon, Google, Microsoft, other major players as well, they're they're softening their customer expectations.

Specific quote here for you is that the hype around or the hype about the technology has outpaced what it can deliver at a reasonable cost.

So what we're seeing here is a shift in both mainstream and alternative media out there about the, the hype around AI. And I'm I'm wondering if we have reached peak of AI hype and now we have maybe even crested it and are now descending quickly into the trough of disillusionment where we reevaluate if there really is value here or if, this indeed is just hype. Now I do believe the application of this technology in the tech world in our industry, gentlemen, does have value because we are talking about the application of machine learning models and mathematical algorithms to data analysis that can help us learn stuff and help help us do stuff.

So I I get that. But taken at a much broader scale and specifically generative AI and and the concept of large language models and interacting with language and that kind of stuff, I do wonder if folks are seeing cracks in the in the in the system, in the hype, and then wondering if there really is value there. I mean, we understand that they're, some of the larger models like OpenAI in particular, they use the entire global Internet as their dataset augmented with with, other data as well. And, folks are wondering, you know, the the presupposition that that is therefore going to yield accurate and excellent answers, maybe that's not true.

Maybe that is actually not, the top quality data.

We're also seeing problems with inadequate risk controls.

We're, of course, looking at the ever increasing cost of training models and building data centers and then, of course, the the, the question of how do we power these things. We even heard of one, organization, building a data center and purchasing a adjacent nuclear power plant and the entire I think the entire purchase was six hundred and fifty million dollars. That doesn't even include all of the hardware to to run these models. And so, I wonder if people are seeing that there is a lack of business value here and, and that the hype cycle has reached its peak, and we are now, going to see a reset in, people's perspective of AI.

I think that they have been sprinkling AI like, you know, fairy dust all over everything where it doesn't need to be, you know, AI on my phone, AI in my search, AI and and when they say AI, they're really not talking about anything beyond the large language model, the the language interface with things. Your AI on my vacuum cleaner. And it just it doesn't need to be there. I think that there is, both at a consumer level and an enterprise level, a an exhaustion about it.

And these early forays haven't added any value to it. You know, the the you know, when you add it to your CRM, when you add it I'm working really hard not to name names. So, you know, I'm I'm trying. You know?

When you add it to your CRM, when you add it to your, you know, browser system, when you add it to your phone and it's, like, there and it's in the way, that leaves an impression of why am I paying for this and how do I stop?

I, you know, I just did a podcast not that long ago with Charlcye Mitchell, and, she really did a great job of out of outlining the real potential of AI in capital letters. You know? The real things it can do besides talk like a human, but not really, but kind of confusingly. You know? It can do all these other things. We haven't gotten there yet because, Phil, to your point, the hype cycle has been so relentless and so facile that you couldn't get past it. So please god, let's get past the marketecture, and let's, you know, let's let the marketers experience the trough of disillusionment so that the technology itself can finally start to be implemented in real, meaningful, useful ways.

Yeah. And when we say technology, let's make a clear distinction here because everything that I read thus far in researching this one article, and then I tried to you know, the there's always links and things you can follow.

They were all referencing generative AI, the use of natural language processing and large language models. Nothing was and and a couple of people were talking about, like, oh, the Centurion AI that's gonna take over. I'm like, okay. That's an allusion to artificial general intelligence, AGI, which no one's talking about here. And I promise you that there are governments like our government, the United States government, and other governments that are racing to be the first ones to have a working AGI.

And and so no. There is there is development. There is money being spent that is definitely something real and not marketing hype. However, like, getting back to generative AI and all of these, you know, assistants in various industries, I I do believe that there is value in certain industries in certain contexts.

I mean, our company uses it, and I think it's fantastic how you can interrogate data using, you know, basically a large language model back end and and a rag pipeline on the front end to be able to decrease your context window to network data. Awesome. Cool. Makes your life easier.

But for a lot of industries out there, the business value is very unclear. So I I think it's clear for a lot of tech companies. Sure.

Especially when you're trying to correlate data and all the cool.

For a lot of businesses, it's unclear, and so it's just trying to find a reason to jam it into your workflow, which ultimately, for a lot of folks working, like, day to day workers, are like, this is actually a burden to me.

This is actually decreasing my productivity because I don't know what I'm doing with this thing, and it's a pain in the neck. Oh, I gotta do this now? I don't know.

So I I think that when you say the trough of disillusionment alluding to the Gartner hype cycle, I think a lot of just people trying to use it are gonna face this trough of disillusionment because the VC money is still flowing big time. Hundreds of billions well, no. Hundreds of millions of dollars.

I was just looking at there there's a couple of websites that I that I check out where you can see, like, what people are doing as far as, you know, VC spend and what they're investing in. And there are several hundred, startups right now back in this year in twenty twenty four, where, we're looking at thirty nine early stage deals at almost nine billion with a b dollars, eighteen eighteen late stage deals valued at over three billion in investment. And that's not the other couple of hundred, you know, angel investment kinda like seed deals, you know, initial investment that are in the hundreds of millions. So money is still flowing.

So I don't know if that's a juxtaposition. I don't know how to interpret that necessarily.

Well, I think we're still early days. Right? I mean, like, if I think back to when SDN first took off, same type of thing. Like, there was this huge amount of hype. SDN's gonna change everything.

Some of that was true. Some of that was marketing. Right? But there are still real use cases that we are using today.

SD WAN exists. There's a lot more orchestration automation in the data center, like, you know, in a way in a manner thinking about a Kubernetes is kinda SDN. Right? Your software define where you're placing your applications and how the networking is built for the interlay on that.

So there are some tangible use cases that can I think AI is gonna be the same way? Right? There's gonna be a lot of hype. There's gonna be a lot of money thrown around talking about AI.

So it's gonna pan out. So it's gonna die off and not not come to fruition. But I I think that, as practitioners, it would be silly to just ignore that AI is just a hype and it's gonna go away and nothing's gonna come of it. I think there's gonna be plenty that comes.

I mean, look at NVIDIA just released their earnings. I think it was last week or the week before. Twenty billion dollars were spent on GPUs for, you know, for AI. So there's plenty of training of AI going on now.

What is the ROI that an end organization is gonna find for that? I think it's they're still trying to figure it out. And I think that's part of, like, if you go look at turnover is one of the hyperscalers, earnings where the market street was kind of disappointed that they didn't have a lot more revenue increase because they've been spending heavily in AI r and d for, like, ninety days for an entire quarter. And like, oh, you don't have more ROI?

It doesn't happen that fast. Right? We all know, like, they're investing in research and trying to figure out where generative AI and some of these models are gonna help make them more efficient, where the sales are gonna come from, and it could take a year or two. Like, it's not just gonna happen in in a, you know, one quarter cycle, in a ninety day cycle.

Yeah. We are talking about a matter of years here. Think about the models that are popular today. OpenAI, Google has, Bard and Gemini and and there's others out there.

The investment on these or for these models, in the training of these models and the acquisition of hardware and maybe repurposing a data center, building a new data center. That happened years ago years ago, and they're only seeing the fruits of it now. And then the fruits are still in question as well. I mean, is there a a return on investment today?

Microsoft, some argue that they're not making any money, on on, their AI product while others will claim that they aren't intending to make money, that they're reinvesting that money and they're just being part of that. I mean, they're investing tens of billions of dollars as our other webscale companies.

And they're investing that money on an eye for tomorrow. And that's very different than VCs looking for a quick return on their investment in the sense of next quarter or several quarters down the road like you said. It's also different than new startups that are leveraging someone else's publicly available a AI. Right?

So, you know, you leverage OpenAI through their their APIs, and then you augment that, with, your own rag pipeline so you can interrogate other contextually relevant data, and then boom, you got an contextually relevant data, and then boom, you got an AI startup. That's very different. And and folks are are, you know, that that's a quicker return on investment, but even there, folks are seeing, is there really any application out there doing that that's making real money? Now adding that as a functionality to an existing class, an existing platform, I mean, that can make sense especially in our industry.

I get it.

But to have that as a standalone application making money that people are gonna pay for doesn't make any sense to me. If you if you think back to the days of the very early browser wars where there was like Netscape, remember that? And then there was what was it Netscape Navigator? And then early Internet Explorer.

You had to pay for Netscape Navigator. I think it was like fifty dollars a year. I I don't remember but it was you had to pay for it. And, and it was arguably a much better browser than Internet Explorer.

Well, lo and behold a few years later Internet Explorer was able to through various deals and things like that have Internet Explorer, as the default browser on various operating systems for free. And so the investment there was, well, this is our loss leader so that way we can reap future rewards. And so it could be that companies are looking at generative generative AI in the same way that it is not necessarily a loss leader, but maybe a loss leader to generate sales in other areas of the business. And so it's something that's incorporated to add functionality, to generate hype, to generate investment money for sure.

But in and of itself, not not really a compelling enough technology to stand on its own.

The beauty of this is that those cloud, models exist so people can experiment a lot more cheaply. Right? They can spin up, use chat, GPT, use Gemini. They can use one of these cloud based models, try and train it, figure it out, see, hey.

Does this actually you know, I'm gonna use McDonald's as an example. Right? Does because they've been in the news about some of their, like, AI and you talk to a robot to order your lunch. Like, does that actually work?

Well, they can start off training that in the cloud. If it works, then they can build a data center and have their own models, their own compute, their own hardware that builds their own models. If it doesn't work, well, I didn't spend a whole bunch of money training that I don't need. Right?

I mean, that's the beauty of cloud is the ability to be able to experiment, spin things up, spin them down. Right.

And that I I was gonna say that using SDN was a really interesting counterpoint because the the real limiter on SDN because I remember when, you know, all the network companies were hyping it and and all that stuff. And I was talking to a buddy of mine who worked at a fairly large medical supply company, and he says he he was at Cisco Live with me, like, we met up there. And I said, what do you think about all this stuff? Like, this is really cool.

He says, sounds great. In another six years when all of my gear has depreciated and I can actually afford to buy some new switches and routers, I might think about it. But right now, I just bought a whole rafting. I'm not getting rid of that stuff.

And that was what limited a lot of people from getting involved in SDN was their gear didn't support it, and they weren't replacing what they had because it was still passing packets. It was still under warranty.

AI doesn't have that. It's almost purely software. Yes. Yes. We talked about how if you're gonna train your own model, you have to build your own data center, and you have to buy your own nuclear reactor to, you know, be able to power it.

And but if you're gonna build a like, you know, your your quick and dirty AI off of chat jeopardy, which is the proper pronunciation, by the way, you know, if you're gonna build it, then then you can do that pretty quickly, you know, in that fail fast model. But it also means that a lot of junk that should never have gotten out of the ideation phase does, and it afflicts us. And that's part of the reason why I'm such a, you know, curmudgeonly skeptic about it is just I'm tired of seeing bad ideas brought to light.

It's fine to fail fast. Just don't fail fast at me.

Yeah. Yeah. I get that. I mean, our our industry, the networking industry specifically, our technology is the substrate for all of service and application delivery.

So we don't want to have, you know, new ideas that haven't been tested and that aren't stable and reliable thrown at us. So that way we're trying to, you know, put shove a a square peg in a round hole and use a tool that may incorporate some sort of new risk into networking that we really don't want. The idea of fail fast and fail often from software development has a lot of merit, but it's there there isn't a direct analog into the networking world where we are much much more risk averse. Now I do believe in the concept of deploying much smaller, configurations that are more easily rolled back and we can do that, more quickly.

That all makes sense. But but what you're saying in terms of, AI startups throwing stuff against the wall and seeing what sticks and then expecting the industry to sort of be their their QA department, and specifically, in our case, the networking industry, that that doesn't just that doesn't fly with me.

Okay. You you know you know the joke about you know the joke about the thing you say after reading a fortune cookie fortune?

Okay.

In test is the IT version of the thing you say for fortune cookies. Right? Is fail fast in test. Do not fail fast in prod.

And I think that people miss that a lot of times is they they think fail fast everywhere all the time. It no. No. Nobody wants you to do that.

Yeah. Yeah. I mean, a very interesting topic to flesh out a lot more on the main show Telemetry Now for sure. Well, gentlemen, I think this is a good spot for us to end, and I do appreciate and thank you so much for joining me on today's episode.

Now for our audience, please remember that today's episode of Telemetry Now was an introduction to our new podcast, Telemetry News Now, in which, I, along with Justin and Leon as cohost, will be bringing you some of our what we consider the more interesting news of the day in the tech world, providing commentary, some witty banter, and hopefully something meaningful to you about what's going on in our industry. Episodes are gonna be biweekly opposite the main show, Telemetry Now, and covering significantly more, news items, as compared to today, which was time constrained due to the fact of it being an introduction for you.

So with that being said, gentlemen, again, a pleasure. Leon, how can folks find you online if they have a question, a comment, or probably a very grave concern about something that you said today?

Absolutely. Because I live I unfortunately open my mouth and end up offending half a, you know, dozen million people.

You can find me on almost all socials at Leon Adato. Blue Sky is where I hang out most right now, although you can find me on Mastodon, LinkedIn, not on that other place.

And, you can also find my blog, adatosystems.com. And, of course, you can find my big deep thoughts, on kentik.com/blog.

You know, we all three of us post there, so you can find my stuff there as well.

Great. Thanks, Leon. And Justin?

Yeah. I am on, LinkedIn. Justin Ryburn, r y b u r n. I am also on Twitter, X, whatever we call it these days, although I don't spend a whole lot of time over there. Do have my own personal blog at ryburn.org, and like Leon said, contribute to the Kentik corporate blog at kentic.com/blog. So one of those ways is probably easiest.

Great. Thanks, Justin. You can find me on Twitter at network_phil. Still pretty active there.

Also very active on LinkedIn. And you can find my blog at networkphil.com. Now if you have an idea for an episode or you'd like to be a guest on the show, please reach out to us at telemetrynow@kentik.com. I'd love to hear from you.

So for now, thanks for listening. Bye bye.