Telemetry News Now.
Welcome to yet another edition of Telemetry News Now. I'm your host Philip Gervasi joined today by my co-host Justin Ryburn. Leon Adato is not with us again this week, but he will be joining us again soon. So make sure to check out future future episodes for the three of us coming back together again.
Now, we are fast approaching Halloween. Next week for us as we're recording is Halloween, Halloween week, whatever you wanna call it, and, this episode you're probably listening to on Halloween or perhaps the day after. Not sure when we're gonna drop it yet, but are there any spooky headlines for you today? I'm not sure.
I'm not sure how to interpret it. I'm looking at my list here. I guess it depends on how you perceive the future of AI. That might be spooky for you.
I don't know. In any case, why don't we get started with our first headline of the day?
Alright. Thanks, Phil. First up is an article from The Register with the title critical hardcoded SolarWinds credentials now exploited in the wild. The article goes on to say hackers have actively exploited login credentials from SolarWinds product called help desk version, Web Help Desk, and it exists in versions before 12.8.3HF1
that are obviously connected to the Internet.
The CVSS score on this one is nine point one.
I think there are a few key takeaways from this article. The most obvious is that hard coding credentials is a really bad idea.
You know, Phil, you've probably had the same experience I have, especially in a lab environment. It's really easy to put a real easy to remember password on something or to hard code it into it when it's in a controlled environment where it's not supposed to be connected to the the Internet or out to the to the broader ecosystem.
But, you know, that's just there's so many ways that can go wrong these days. It's just a really dangerous and really bad idea. And then I think the second lesson here is to keep these kind of things in mind when you're migrating applications to the cloud. If you're doing lift and shift, if you built the application and designed it to be behind a perimeter firewall and not have access to the Internet, not be accessed from the Internet, if you migrate to the cloud and have it have access to the Internet, that may expose risks you didn't have before.
You know, and I think the the final lesson here is the ever present reminder to patch early and patch often. So SolarWinds does have a patch for this.
You need to upgrade your web help desk to twelve point eight point three h f two.
All of the known exploits are to versions that are unpatched. So patching is your friend in this case.
Yeah. Patching is definitely your friend in this case and and and always. Right? But, like, nobody ever wants to do it.
Or it's not that no one ever wants to do it, but it just becomes one of those, like, mundane things that that I think, that it gets lost in the routine. And so or or maybe it gets pushed down to a lower priority, and so it doesn't get added to the next change window or or things like that. And so I think that's probably the issue here is that, you know, it's part of the it's part of the routine of network operations, especially for a solution that is so heavily integrated into your entire infrastructure, and, and therefore, it's critical. And and, yeah, that's a really high CVSS score.
So if you're out there dealing with this, you know, good luck to you, and, and I hope you don't experience any severe issues as a result.
Yeah. Either that or remove your the Internet access to your your web help desk software.
Yep. Absolutely. Absolutely. So moving on to other news, a story out of Reuters last week, week and a half ago now, NVIDIA briefly surpassed Apple as the world's most valuable company, and that was on October twenty fifth with its market value reaching three point five three trillion dollars driven by a strong demand for its AI chips. Now NVIDIA closed the day last week with a valuation of actually three point four seven trillion, which is slightly below Apple's current evaluation of three point five two trillion, but it has been maintaining significant growth momentum, again, fueled presumably by the global boom in AI and, and also a recent investment boost from OpenAI, which is very interesting.
And if you may recall, in twenty twenty three last year, NVIDIA stocks surged almost two hundred percent over the course of the year. A lot of instant millionaires overnight that we saw all over social media in the news. And that was predominantly, fueled from, the the rising AI adoption across the tech industry, but a variety of industries. And, and so the the company, NVIDIA's dominance in AI processors, it really does place it at the forefront of a competitive field right now, with other tech giants, but including, you know, those web scale tech giants like Microsoft and Alphabet as well and now Apple. So a lot of interesting stuff, a lot of talk about, future options, and so there's a lot of popularity among options traders, not just the tech world, and, and really cementing NVIDIA's influence, of course, on tech, but also really on the broader US stock market.
Yeah. I think in a previous episode, I talked about how, I had read a book called Chip War, and it talked about, you you know, kind of the history of Philip and all the geopoliticals around that and stuff. And I everybody thinks of and talks about NVIDIA's chips when they talk about how well they're doing financially. Right?
And that is definitely a part of it. It's a big part of it. The thing I underappreciated until recently, I was listening to a podcast with, the NVIDIA CEO, Jensen Long, and he was talking about their actual broader strategy. Right?
They're actually building a moat around their product, and they're building an entire ecosystem, everything from the software that train that spins up all the GPUs, gets them programmed ready to go, does all the networking, monitors all the networking.
I mean, they've got a holistic ecosystem around their AI compute data center environments that's not just the GPUs. That's where a lot of their money comes from, but they're building this kind of holistic ecosystem. So even if AMD and Intel catch up, which I fully expect they will when it comes to the design and the manufacturing of the Philip, the GPUs Yeah.
It's gonna be hard for them to catch up with the ecosystem that NVIDIA has built.
So I think this will be you know, are we gonna see some slowdown in AI spend? Sure. It's a possibility, but I don't think that's gonna make a huge impact in NVIDIA's earn earnings over the next, I don't know, three to five years. They've still got a long runway, I think.
Yeah. And, certainly, they are, also benefiting from the speculative investment as well, considering that they have been the poster child for the chip manufacturer that you wanna invest in, that's supplying the chips that's making this I AI boom happen. So so folks thinking about, I wanna get into this market. Who do I who do I put my money toward? And it's gonna be NVIDIA.
Alright.
Next up is an article from CRN titled Upstart Alkira launches ZTNA, which stands for Zero Trust Network Access as part of single cloud networking platform.
Alkira is a network infrastructure as a service networking platform.
The article states Alkira ZTNA supports complicated multi cloud environments from a single portal and tightly ties together networking and security operations.
The service can also integrate with third party security vendors such as Palo Alto Networks and Fortinet.
Interestingly enough, I think mostly of Alkira for their cloud overlay and how easy they make it to deploy application workloads in different cloud environments without having to make sure you get the networking pieces right, each one of the cloud environments, because each do it slightly different. But this article actually really focuses on more of their traditional NAS or CASB type of, setup where they have pops from around the world, what they call their cloud exchange points.
They have a client that you can run on, you know, an endpoint. So if you're not in a a branch office, if you're working from a coffee shop, working from home, they have a software package called Secure Connect Client that allows you to get access. So I I really hadn't, I guess, been paying enough attention to their product product portfolio. They've got some additional pieces that they bring to the to the table here.
Yeah. Yeah. And I I think this additional piece is just speaking to that, desire or seems like to me at least their their goal is to make, network operations in the cloud easier. And so they are very operational operationally focused.
Although you did mention that they have that backbone infrastructure, and I get that. But, this, I'm I guess you could say was kind of the missing piece. Bringing the ZTNA piece into it, allows engineers to that are already using Alkira with minimal cost and with minimal friction incorporate that security stack into their cloud environment. And and that's especially if you're considering moving away from, you know, funneling everything through some virtual firewall in a centralized security stack and then going toward a host to host or app to app security model, which is, for a lot of people, the way things are going.
And it's also a lot more difficult to, implement and manage for engineers just in, you know, a daily operational perspective. So I think that this is a great move. This is a this is a it's right in line with just progressing with where the industry is going as well. So, you know, I'm glad to see that from Alkira.
And they're and they're good friends of ours here at Kentik, so we appreciate what they're doing. And now it's time to talk about nuclear power. That's right, folks. Here at Telemetry News now, we bring you news about submarine telecommunication cables and satellites in outer space and today, nuclear power.
Specifically, Amazon has signed new agreements to develop small modular reactors, SMRs, as part of its strategy to transition to carbon free energy and meet its climate pledge of net zero carbon emissions by twenty forty. So Amazon, which already matches its energy use with a hundred percent renewable sources, it's aiming to address this growing energy demand by investing specifically in nuclear power now, which by many accounts is a scalable carbon free source of energy with a proven safety track record.
Now right now, they're in partnership with Energy Northwest, and, Amazon is gonna support the development of four SMRs in Washington. That's gonna generate, an expected nine hundred and sixty megawatts. That's a lot of power. That's by the early twenty thirties.
And they're also working with Dominion Energy in Virginia to explore another SMR project that should bring about three hundred megawatts to meet local power demands. So there is a lot of activity. There's a lot of interest in nuclear power, especially with the advent of AI data centers. So it's very interesting to see this investment, this change in the, in the conversation as well.
And, not to mention just the the the new focus on power, which is something that we really kind of take for granted and haven't talked about until recent days. You know?
Yeah. I mean, it's interesting you and I have both sat through presentation where people are talking about building out modern AI data centers and the amount of power footprint that one of those data centers can have, not just from the GPUs, but the networking, the storage, everything else that goes around Right. You know, the GPUs.
I mean, they're getting creative and even reducing how how many optics they have and going back using copper cabling because it reduces the amount of power draw to power the lights and lasers and stuff. So, yeah, I mean, all options, I think, are on the table for companies that are building these huge AI workload and Yeah. Data center environments because where do you get the power? Right? I mean, we spun up enough, you know, coal power plants or, you know, enough natural gas power plants to power AI, it would absolutely crush our environment.
So Yeah.
And I'm totally I'm totally reading between the lines here because the these, the announcement from Amazon and then from subsequent news articles were about their their pledge to be, carbon friendly or, you know, climate friendly and carbon neutral and all that. But, we both know that this is, a push to to, source and secure high quality, consistent power for that kind of activity. That's a hundred percent what it is.
Yep.
I mean, not that not that, what they do already doesn't require significant amounts of energy. And, of course, they're they're trying to grow like every other business, but, certainly, I think that it is, spurred by and large by the growth of bespoke AI data centers, for sure.
Yeah. And like you said you alluded, we've we've unfortunately had some really bad natural no. Not natural. Human made disasters when it comes to nuclear cores melting down and so forth, which, you know, the but the technology has come a long way since those, disasters took place. So it'll be interesting to see if we can build nuclear reactors that are smaller, that are, you know, safe, that have some safeguards in place from those type of things happening.
You know, if we can, it's a much cleaner way to get power than some of the other options we have on the table, so definitely worth keeping an eye on. Yep.
Alright, the next article is from computing, I'm sorry, network computing titled SpaceX files with FCC to offer gigabit services. The article states satellite internet service provider SpaceX filed papers earlier this month with the FCC to approve of changes that would enable its second generation satellites to provide promised gigabit services to all interested parties.
If approved, the orbit of the satellites would be changed and service delivery could begin. The company's Starlink Internet service, which is used worldwide to connect hard to reach and underserved areas as well as in the wake of devastating natural disasters and in war zones, provides users with increased network resiliency and a backup option to terrestrial networks.
You know, I think I I said previously on the podcast, I find this kind of thing really fascinating because Mhmm. You know, in twenty twenty four, access to Internet, we kinda take it for granted in the developed world and in places where there hasn't been a natural disaster. Like, it's it's how we communicate. It's how, I mean, COVID proved anything.
It was that that's how our society continues in a in a pandemic like that, right, is the Internet. I mean, without it, like, can't communicate, we really can't live our lives. So, you know, I'm really excited to see satellite Internet in general being able to fill in some of the gaps in some of these hard to reach places and especially in war zones and natural disaster zones and being able to take it to the next level and offer gigabit speeds is just absolutely amazing to me. Yeah.
Yeah. I agree totally. And and, frankly, I've been amazed and and impressed with what Starlink has done up until this point aside from gigabit connectivity. Just getting a couple hundred megs with low latency or reasonable latency anywhere in the world via low earth orbit satellite is phenomenal. I mean, like you said, our entire world, for the vast majority of people, connectivity to the Internet and access to information is critical, whether it's for mundane stuff, communicating with others, doing business, learning, education, all these kind of things, participating in government, local, and and at a broader scales.
And so you know? And and that's all possible to do with a couple hundred megs. I mean, I have gigabit to my house, and I barely scratched the surface of that. So to be able to offer very high bandwidth Internet, and I consider gigabit, high bandwidth, literally anywhere in the world, I think, is a huge step forward in just, further providing a mechanism for all people to be connected and to engage in those things that our society values right now, especially as we start to see this technology get smaller and then maybe even integrated into our phones.
I keep waiting for an announcement about that. That's gonna be exciting, to have that kind of connectivity, literally anywhere in the palm of your hand. So really cool and, love seeing news coming out of Starlink. And I guess my only complaint is that I want more bandwidth.
Like, one gig is not enough. I want it all. I want all of the bandwidth.
Well, the article does mention that Musk claims that they've tested up to eight gig. So, you know, that's not what they're asking for approval of or what they're planning on rolling out today, but supposedly, you know, just sit tight. It's it's coming. But, yeah, I mean, the low Earth orbit is really interesting to me. Right? Because, I mean, satellite Internet has been around for a long time. Sure.
You know, satellite Internet in and of itself is not a new thing.
Like garbage.
Problem was yeah. The problem was to get the coverage that most satellite companies were offering prior to Starlink, you had to be so high up in the Earth orbit to be able to get the reach, right, to be able to get from pole to pole. Like you said earlier, that your latency was so high that even if you could have good bandwidth, your TCP windowing on your laptop when you go to do anything is gonna be terrible. Right? Just keep in mind the throughput that you get, the experience you have as an end user is factored by both your latency and your bandwidth. We always talk about bandwidth, but latency is equally important.
And so where lower earth orbit brings Yeah.
Is lower latency.
Mhmm.
So, presumably, they are gonna redo where these satellites are orbiting, but, presumably, it's still gonna be lower. So you're still gonna have pretty good, you know, latency within some margin of error with these high bandwidths.
It'll be interesting to see what kind of, you know, performance that gives an end user.
But, I'm definitely excited to see where this goes.
Have you seen the, like, trail of Starlink satellites going, like, over your house? Or Mhmm. For me, because I live in New York, it's like when I'm camping somewhere, not when I'm at home. Of course.
Yeah. But it's pretty neat to see them, like, all in the line. I remember the first time I saw it. I think it was over the summer or maybe last spring.
And and I and I I immediately whipped out my SkyView app on my phone, and I'm like, what is that?
It did not show up on the app, but immediately, I, I realized that's Starlink. It's just really neat to see that and, really cool technology.
So Now speaking of the new thing now experience.
What's up?
Sorry. I was gonna say I had a same experience where we were out camping, and, I was inside the camper. My wife yelled to me. I came outside, and she's like, look.
There's aliens invading. I'm like, what is that? I had to get it figured out. Now they know what it is.
It's not nearly as scary.
What you said No.
Not at all.
Of what looked to be stars all tracking together in a line. It's like, that doesn't look natural.
Yeah. Yeah. But still awesome. Right? I mean, I love going outside, especially with the kids, and you see the line of satellites or you see the International Space Station or something like that.
Very, very cool. So moving on to other cool technology. No. Not satellites and not outer space, but artificial intelligence.
Anthropic has launched Claude three point five SONNET, the first model in its Claude three point five family designed to deliver advanced AI capabilities.
And they're, claiming that they have and I've seen some of the benchmarks in the research, improved speed, improved intelligence, whatever that means, and we can get into that, improved cost efficiency, and some other benefits as well. Now if you're not familiar with what Anthropic is doing, they sort of fill a niche here in kinda like the mid market. I I think Claude and the Claude family of models are are kinda like a mid tier, large language model, and that's, as compared to, like, the the very largest foundational models like GPT and then very, very small large language models, which some people even call small language models, like, like Mistral or something like that, or even smaller than that, really, because there are some iterations of llama that are just a few hundred, five hundred million parameters as opposed to hundreds of billions of parameters or even trillions.
And Claude fits right in the middle there, and and therefore is a great option for folks that need that enterprise level, capacity, but not necessarily the cost or or the, the complexity maybe of operating a larger foundational model like GPT. And so some of the things that folks are getting excited about with Claude three point five Sonnet specifically are it's, what they call visual reasoning. Right? Enhanced vision capabilities.
And that that's like interpreting charts and graphs, transcribing text from kind of, like, mediocre images, which, in my opinion, makes it very, very useful for for more real world world scenarios, I think, where, you know, you have a lot of visual input and it's not perfect.
They have a greater context window of two hundred thousand tokens. I don't know if previous versions were I think they were at around two hundred thousand tokens as well. But in any case, so they're increasing the size of the context window, which may not be relevant for you because the the larger the context window and the more you dump in the prompt, you you can experience inaccuracies and hallucinations, because of a a phenomena called lost in the middle. Very interesting.
Look it up. There's efficiency improvements, cost improvements. So right now, with claw three point five SONNET, the price is three dollars per million input tokens and fifteen dollars per million output tokens. Oh, it's So something to consider as well.
Now something that I wanted to mention here is that I I really like seeing this, these iterations of models, just overall improving in accuracy, focused on speed and performance on real world use cases and scenarios, but also very much among the smaller models. It's not just the the GPTs of the world that we're talking about, but much, much smaller models that are making improvements and in some in some use cases, even outperforming the very largest models. So really neat stuff going on. Now Claude is not a free and open model in the same sense that, like, Llama is, but certainly a great option and something to explore, moving forward, especially if you're starting to integrate large language models into your own, like, network operations or IT operations workflows.
And, I I mean, I love how everything is just becoming more accessible right now. It's just easier to spin up models, even even larger models, and you can run local models locally. There's just so it it's just getting easier, the resources available online.
Of course, if you're learning it for the first time, the tutorials and the blog posts and the videos, all of those things are just kind of flooding out right now. Really exciting, and it's, in my opinion, very easy to get started up and running right away. So really cool stuff.
Yeah. That does also scare me at the same time. Right? The the easier you make it accessible, the easier it is to get it in the hands of some someone who's trying to use it for nefarious purposes, right, as a, you know, an attack vector or whatever.
So there's also that. Although, I will say, you know, what little reading I have done on the various different, I guess, vendors that are developing these AI models, and turns out it's not just OpenAI. Right? We hear about them every day, but Anthropic is out there.
There are a number of other companies that are that are building these models. And Anthropic seems to be investing quite a bit of time and energy into thinking through how to do it safely and working with some of the law enforcement agencies, trying to figure out what good policy looks like. So, you know, they are trying to consider that angle. I'm not saying the others aren't, but there is some amount that the industry is trying to think through, like, what, you know, what are the downsides to the technology we're in we're inventing here.
Right?
Well, I mean, I'm I'm sure there are some that folks can speak of right now, and, it certainly opens up new attack vectors for for bad actors. But I don't think that precludes that we still develop and look into and research new technology. Right? I mean, technology in my opinion, technology is is amoral in the sense that it's it's a thing.
It's not a person, like, with a soul that has, an understanding of right and wrong. Right? Now where that's gonna go with a AGI and stuff, I don't know. That's not that's not the point I'm making here.
But in the sense that it's it's just a technology, and it's really how do we use it, I think that's the concern. And I and I I know you agree with me.
Oh, yeah.
Yeah. And so I think I think, there are some some concerns, data privacy, new attack vectors that folks are starting to take into consideration now. So for example, there are folks that I'm I'm talking to and working with that instead of using some of the very, largest publicly available models, they use a much smaller model locally on their laptop or or whatever infrastructure, they're running locally. And in that sense, they're not connected to the Internet literally and and sending that information. They're just using the, the compiled code of the model to run whatever operations they wanna run locally. So so folks are figuring out what to do with it.
And and, sure, there's there's always needs to be a healthy skepticism. I'm not gonna say fear necessarily, but, you know, precautions that we need to take as we develop new technologies. So, yeah, I agree with you there. And now moving on to upcoming events over the next month, six weeks or so.
We have the Ohio networking user group, ONUG, in Cleveland on November seventh, so in about a week. We also have the Texas networking user group. This one in Houston, also on November seventh. If you're not familiar with the NUGS, the networking user groups, please go check out the USNUA site, USNUA dot com.
Click on groups and events, and you can find an event near you.
NFD thirty six, November six and seven is gonna be live streaming. So whether you watch the live stream or you watch the, the videos on their YouTube channel after the fact, I encourage you to check that out and see what, what's going on in the, coming out of Silicon Valley in the world of networking.
We have KubeCon, November twelve through fifteen. Justin, are you gonna be at KubeCon this year? I am. Oh, excellent.
Excellent. So if you are too, make sure you go say hi to Justin. I will not be. Microsoft Ignite, November nineteen through twenty one.
And, AutoCon two, from the network automation forum is gonna be November twenty through twenty two in Denver. And, I think we mentioned it last week, but I'll mention it again.
Myself, Justin, Steve Meuse, and Mike Krygeris, our colleagues here at Kentik, will be doing a workshop on network observability, and I'll be doing a talk on LLMs and network operations.
And last but not least, one of the bigger events of the year, AWS re Invent in Las Vegas this December six rather December two through six.
I'm not going to AWS re Invent, though, Justin. Are you going to that one?
I am going. Yeah. I'll be at that one this year as well. It's gonna be a busy, event season between now and the end of the year for sure.
Yep. Yep. This is the busy time for tech people. Lots and lots of events, lots of conferences, and, which is actually why I chose to abstain from attending re Invent.
I was on the list to go and even got the email reminder to book my travel, but just so many things going on. So I will miss seeing you and others, in person, though. That's that's for sure. And those are the headlines for today.