Telemetry News Now

Telemetry News Now.

Welcome to another episode of Telemetry News Now. We are recording today on Wednesday, August six two thousand twenty five. I almost said two thousand five. My goodness.

And we do have a bunch of headlines for you today, but I do wanna start with some significant news from the Gervasi house. Justin, I don't know what the basement situation is in your home where you live outside Saint Louis, but here in upstate New York, I do have a below grade basement. And it's very humid in the northeast in the summer, so I do run a dehumidifier pretty much all summer. Well, it's a dry basement, but I do run one.

And I've always had the task to manually remove and empty the water container, the bucket every day. And the only reason I didn't attach a hose is because I'm lazy. And finally, I made a trip to Home Depot, picked up appropriate length hose, and then ran a hose from the dehumidifier to the corner of the basement where I have the sump pump. And I'll tell you, it's the little things because it that is a game changer in my life that I don't have to think about it.

And it just runs, permanently until the motor gives out.

I think that's what they call automation, Phil. Automated that fast. It was manual and and cumbersome. So yeah. No. Thankfully, don't have to do that in the house that I'm in now, but I have lived in houses in Missouri where dehumidifier was a fact of life. Being able to have it drain out automatically and not have to remove and empty that bucket every time it fills up, I can understand why that's a laborious task.

Yeah. Yeah. And I have an older house where I don't have a humidifier or dehumidifier built into my air conditioning and heating system.

Yeah.

I do have baseboard radiant heat and then a central air separate. So so I know some newer houses have that, but not me. But thank you for bringing it back and bringing it home and somehow finding a way to tie that to networking. Very good.

So let let's get into the headlines today. We're gonna start with some big news from Intel. Now you may have heard of some of the layoffs happening, but from CRM and and other sources last week, Intel plans to spin off its Network and Edge Group, also known as NEX or NEX, into a standalone company. Also seek outside investors, not sure what that's gonna look like yet, and keep a meaningful anchor, quote, unquote, anchor stake, which pretty much mirrors the Altera playbook from earlier this year.

So this this new spun off unit is gonna focus on silicon for communications, enterprise networking, and Ethernet connectivity. So we're talking about actual network cards. And I I haven't seen anything yet about timing, and so I'm pretty sure that the timing wasn't disclosed yet. Justin, I don't know if you kinda looked around and saw anything, but so far, haven't seen anything about that.

Now this move stands or rather lands alongside a broader restructuring under the new CEO, Lipa Pooh Thanh, which as many people are painfully aware, included substantial workforce cuts, thousands and thousands of people laid off recently from Intel. And that's, you know, ultimately to narrow Intel's focus on core CPU, foundry adjacent AI silicon, that's the thing lately, and also monetizing noncore assets. So, you know, I think if this spin off works as intended as an independent organization, NEX could move quicker, obviously, on roadmaps. Right now they're working on the e eight thirty Ethernet family of network adapters, so that's something.

And that's a space where a lot of folks believe that Intel actually trails NVIDIA and Broadcom in high end throughput, high end offload. So that could be a net gain for that initiative.

And I think if this carve out accelerates Ethernet innovation, right, I mean, that's good news for for network operators, especially if you're betting on Ethernet based AI fabrics, which you should be. But, you know and I think, you know, you know business, Justin. Until funding leadership and then there's like a concrete road map are public, I'd still treat this as like a watch and verify situation. So in other words, you know, lock in your supply commitments and kinda keep a plan b for your Nick vendor and your bills of materials.

Yeah. I mean, like you said, the timing isn't fully disclosed. They did the articles that I was reading did say that they still haven't finalized the sale of Alterra to Silver Lake. That's supposed to happen in September.

So presumably, they're going to have to bring in a lot of times what you'll do is you'll get a contract with an investment banker, right, that will come in, especially if they're going to, like, make this a publicly traded company that they just hold a percentage of the shares in. They'll need to come in and get an investment banker. They'll come in, do due diligence, figure out what the asset is worth, figure out how to price the shares, help them do the offering and listing. So there's a lot of, you know, finance type of work that has to be done before they can, you know, list it.

If that's what they're going to do is what really isn't clear whether they're going to publicly list it or they're gonna try and find, like, private equity like they did with Altera where they're bringing someone like Silver Lake to buy Yeah. The other parts of it. So it's unclear yet who the whether it'll be publicly listed or be a private sale, but, yeah, definitely, it'll be interesting to see where this goes. And presumably, this helps get some cash back into Intel parent company, right, which helps them accelerate their roadmap because as we know, they're they've they've fallen behind when it comes to some of the GPUs and stuff with NVIDIA.

Yeah. Right? Yeah. So presumably selling off some of these assets which still have some value.

Right? It allows them to get some cash that they could infuse back into the business and and build faster road maps, try and catch up with NVIDIA on some of the GPU stuff.

Yep. Yep. I did see a couple of mentions of the possibility of PE firms involved. And, you know, it's not, I think, outside the realm of possibility for even Intel to be one of the owners, Intel proper.

Right? And not as a subsidiary. Right? Not as but just as another investor and owner.

So they have some some stakes still retaining. And, you know, I agree with you about the networking component and and this ability to be more nimble and agile. Not necessarily nimble and agile, but be able to innovate more quickly and focused so this particular organization, this the whole networking organization can be focused in a way that it necessarily it might not necessarily be able to right now. And, you know, I, you know, I foresee the possibility of some partnerships with OEMs and cloud vendors that we haven't seen before or at least strengthening and making that a little more of an aggressive initiative.

So I think there's some good opportunity here.

I wasn't expecting that, and certainly, you know, on the heels of the workforce reduction that was very significant and top of mind for a lot of folks.

Well, had to think that I mean, they didn't bring in a new CEO to keep things status quo. Right? So you had to know some things were coming. I I figured.

I didn't see this necessarily coming. You know, unfortunately, the layoffs have have been announced. We knew that was coming. And I'm sure there will be, you know, more changes coming.

Right? I don't know what they look like, but they brought in a new CEO to kinda right the tide here with Intel. So I expect we'll see more changes as, you know, the longer he's on the job.

So Mhmm. Mhmm.

So from Network World on August five, Cisco's Foundation AI team is partnering with Hugging Face to scan every public file uploaded to the platform, its models and and other AI related artifacts, with an upgraded ClamAV engine that that flags AI specific threats. So think about things like deserialization risks, like in specific files, threats like that, and also do that in milliseconds. So Clam AB's detections that would allow or rather, it would also show up in virus total and also builds on prior joint efforts like Cerberus. The idea here is that it's feeding standardized AI supply chain threat intel and it's going right into Cisco security products. So the integrations then are gonna span things like secure endpoints, secure email, secure access.

And the idea is that you have this ability for policy enforcement like blocking AI models, maybe non approved sources of those models, compliance checks for licensing and things like that. From the article, Cisco says that they're positioning this as quote, democratizing AI model anti malware unquote, and then also scales protection across Hugging Face.

Because remember, Hugging Face is currently hosting almost two million models Mhmm.

And all sorts of derivatives. Not only that, but it's a forum and a community. It's very, very valuable in a lot of ways.

But as far as models, it's almost two million. And and and, Justin, I remember doing a telemetry news now episode with you not that long ago where the number was was like one million. Mhmm. It's just crazy.

Now to be fair, of those two million models, many are variations of the same model. It's not two million distinct foundation models. It's not like that. But, certainly, there's still a lot going on.

And, I mean, security issues, specifically around operationalizing AI, actually deploying AI into your production environment, that's become top of mind for a lot of IT leaders. Yeah. So I think this is a big move for Cisco. And I also really think it's interesting that they're partnering with Hugging Face, kinda like an open AI not open AI, the company, but, like, open artificial intelligence repository and forum and community.

I mean, can get into the security aspects around foundation models like GPT and Claude. That's kind of a different thing. But so many people are using models that they get from Hugging Face right now. Not just in their labs and POC like me, but in production as well.

So I do like how this pretty much allows network security people to sleep at night. You can treat models like artifacts, like any other kind of risky payload. You set up your policies, allow lists, compliance checks, whatever you're doing, right?

Reputation checks for all those model downloads moving over your corporate network. It's really it's really addressing that major growing concern among a lot of IT folks, like security concern.

Yeah. I mean, to your point about how quickly Hugging Face is growing in the Network World article, it said Hugging Face adds a new model on average every seven seconds.

So, I mean, you know, it's like kind of the YouTube of AI models. Right? I mean, they're just exploding right now as far as the amount that are being developed and and hosted up there. And you have to imagine, not all of those two million models that are out there are malware free.

Right? Somewhere, somebody, some bad actor has put malware in one of these models or is training the data on bad data. I mean, there's all kinds attack vectors that, you know, open source could potentially bring when you talk about open model development. And so the the idea here that I from what I understand is that Cisco is able if you've got, like, Cisco firewalls in your environment, is able to actually do some inspection of the a traffic that's flowing through the firewall, detect that there's malware, and then even do the enforcement to to block that traffic.

So it's, you know, not at the network layer, it's actually the application layer, but the AI model and the transmission of the data as the workload workload that it's actually inspecting, which is, you know, pretty cool. I mean, we've had application firewalls for a while for other applications, but specifically for AI models and AI workloads. Kind of an interesting spin on it.

Yep. Yep. And and I would say a necessary spin for sure. So Cisco's getting ahead of this and and partnering with an organization like Hugging Face, think, is a big deal.

So also from Network World on August five, a report from ISG finds SD WAN is now mainstream. That's not news.

That's not the news part.

Welcome to twenty twenty common.

Yeah. Right. Right. And, you know, the report finds it is very common in large enterprises and growing among SMBs.

Not news. You know, I saw that but I saw that happening before my eyes years ago when I Mhmm. Left the VAR space. That was only a few years ago, actually.

And I remember whenever I spoke with customers, and I was networking focused as far as my my specialty. Whenever I spoke to customers about a router refresh or some sort of a WAN project and typically very large customers, Usually, it was an SD WAN conversation. There were exceptions, of course, but what's interesting here in this report is that most organizations are overlaying SD WAN on existing networks rather than doing a full rip and replace.

And so the idea of a fully autonomous WAN with extensive policy, multiple links, all that stuff is by and large, according to the report, largely still aspirational for a lot of organizations and still requires engineers to very actively manage.

I got a lot of opinions here, but I'm gonna keep going. So again from the report, SASE adoption, which does vary by region, by industry, the size of your organization, and also Edge Plus Private G is helping to move this along from an like kind of an aspirational thing to reality. And then looking ahead, ISG expects deeper integration across new transport. So think of like satellite, like Starlink and Kuiper, and also, you know, new WiFi standards coming out.

Interesting.

You know, yeah, I I don't the initial part of this report, I don't think there's anything groundbreaking here because most of the customers that I worked with, large customers we're talking about, unless it was pure greenfield or some kind of interesting exception, they were never really able to do a huge WAN rip and replace from traditional routers in a brownfield environment to a complete pure SD WAN, you know, as a project, as a cutover on Friday night.

I mean, typically, yeah, if you're thinking about a large organization with many, many sites and various types of sites, including data centers and all this kind of thing, you ended up with a three year project to slowly migrate Or more.

Or more. And most of that time, you had a hybrid network. You had many sites where you were just getting the the boxes into place and operating over what you had until you could get circuits moved because, know, you have to put in an order for a new hundred gig circuit at your data center, and that took six months to to get put. You know, the all of those factors involved there.

So, yeah, I do I do believe that it depends on the industry and the size of your organization, that sort of thing. But I also think that a lot of it is, you know, the culture. How much is your security team, your network managers, that kind of thing? How how much are they willing to allow SD WAN policy purely to manage your WAN as opposed to maybe a light touch management?

The answer is, you know, not many. I mean, most folks are gonna want some kind of app operational light touch management even over a fully autonomous WAN. And when I say fully autonomous, I say that almost like tongue in cheek because it's never fully autonomous. You're still actively managing those policies and making adjustments as your network changes and stuff.

But, yeah, I wanted to bring this article into it because, you know, we're still talking about SD WAN, and I I I just sort of had an interesting not that I'm calling my own takes interesting, but I had my own pointed opinion about what ISG found in their report.

Yeah. I mean, we worked at Kentik with a partnership with an SD WAN vendor. This is going back, I don't know, five, six years ago. They actually approached us and wanted to partner with us.

And at the time, I thought, you know, that's kinda strange because all these SD WAN vendors have their own observability of the traffic that flows through their SD WAN. Right? That is you as a as an end user can log in and see all of your traffic and where it's flowing and all that kind of stuff. So I thought it was interesting that they would approach us and want to partner on network observability.

More we started talking to them, what they were tell saying was exactly this, Phil. Right? The majority of their customers that are of any scale or size at all have a migration plan over three years. In fact, the one beta customer that we work on this partnership with was a five year plan, five year rollout.

Right? Because they have thousands of sites around the globe. Right? And not all of those are, like you were saying, within the reach of a nice one gig fiber connection.

Some of them are, like, have to be connected by satellite or the best they can get is DSL connectivity, you know, these days or whatever. So, you know, connectivity is a challenge. Getting the boxes out there, installed, cut over, all of that is a challenge. And it takes time, and it takes a whole migration plan.

So you like you said, most of these enterprises wind up in a hybrid WAN. I don't know if that's a Gartner approved term or not, but, you know, hybrid WAN where some of the sites are on their historical WAN that was presumably like some sort of MPLS service that they were running or that they were buying from a from a provider over to Internet circuits with, you know, SD WAN appliances and an SD WAN solution running on top of it. That migration takes multiple years for a large organization. So that part of it you know, at the time, all the SD WAN vendors had marketing that would make you think, oh, yeah.

You yesterday, you had your old WAN, and tomorrow, you'll have SD WAN, and you'll do the cutover over a weekend, and you'll be good to go or whatever. Well, reality that enterprises are seeing is it's just like migrating to the cloud. Right? For most organizations, unless you're a really small organization, it's not happening overnight.

Mhmm. Mhmm.

And there are some there are some things that we can say with regard to cloud, cloud connectivity. There are absolute major significant benefits that happen when you migrate to an SD WAN even in a hybrid situation. You do gain that single portal to manage many of your sites. You do have the ability to connect to cloud, you know, or whatever VPN connectivity you use much more easily.

There's policy involved. So there's there's a lot of benefit, but there are just so many permutations and aspects to migrating all of your stuff that I mean, just think about when your next change window is. Imagine a change window for your active data center and let's say your data centers are really big because you're a big company, right? A huge national bank.

You're not changing. You know, you're not just gonna buy those boxes, stick them in the rack, and then you change windows this Friday. It's probably like months from now and and series of cab meetings and all these things and rollback procedures and all that stuff.

Testing it out.

All of it. And so and so the reality is that these are very long term strategies. And so again, that's why this report kinda struck me because I was like, of course, it's aspirational for a lot of folks because we're not there yet. Know?

If you're starting your SD WAN strategy just now, it's gonna be a few years before you get there. Again, unless you're greenfield or you have some exception or you can stay you know, I I have done projects where I stood up parallel networks. We were able to do that. Like, all all of the circuits were net new and we were to set up a parallel network, and then we cut things over by literally moving wires or or just changing a route or whatever we had to do, suppress a route and advertise different prefixes, or whatever we did with an easy rollback.

And so you can cut certain things over, like, in a in a night, but it did take us many, many months to get there, of course, in order to stage that entire parallel network and and get all of the policies perfect so that way, you know, your latency requirements were being met, your SLAs were being met, your security things were being met, your even things like certificate management among all your your boxes. And I I remember back when it was a matter of folks coming to me and saying, oh, but I thought I could just upgrade, like, my ISR to the SD WAN image. And I'm like, I I yeah. We're gonna have to like reboot all your routers, install you know, obviously, you install the new image, reboot your router, do the whole spiel, and then there's gonna be a rollback plan.

So I mean, even that was like, I was always really nervous. I'm like, can we just buy new stuff, please? Yeah. So there's there's a lot there.

Back out procedure on that's really messy if that goes wrong.

It it is. It is. It is. And I and I've been there, unfortunately.

But I do think, again, going back to cloud, SASE, things like that, you know, that that does make sense and I believe that the ISG report does call that out correctly, that that is gonna be a major motivator and it is a major benefit as well. So, we're still seeing growth and changes and even just pure adoption for the very first time of SD WAN in spite of it being a very mature technology right now.

Mhmm. You know, the other interesting thing in the article was they were talking about how AI copilots on top of these SD WAN solutions come into play. And they actually made the point that they don't see evidence of completely autonomous WAN tuning using an an Gen AI Copilot. Right?

What they really see happening is more of a supervised learning where there's a human still in the loop. The AI can make some suggestions on rerouting of traffic, presumably collecting all the telemetry and saying, hey, you've got an application performance issue because this application is taking the wrong path on your WAN and can provide some suggested remediations to some of those things. But, you know, their research was showing that customers are not quite to the point where they're doing this completely autonomous and it is just completely self driving, if you will. So which kind of meets with the what I hear from talking to customers, and I I assume it's probably the same for you, Phil.

Yep. Yep. Absolutely.

The next article is from Network World titled Broadcom ships new Jericho four Ethernet fabric router.

Broadcom is adding to their list of chips, the Jericho four chip, which joins their Tomahawk Ultra and Tomahawk six, which are chips that they sell to people who are building network switches, Ethernet switches for mostly for HPC and distributed AI environments. Right? So where people are building AI GPU clusters, distributed clusters. There's that's primarily where these Jericho four chips are being sold into that they're announcing.

Interestingly, that they were saying that the new Jericho four has a lot higher buffers than a lot of the previous generations, which allows for it to do what they call high bandwidth memory packet buffering, HBM packet buffering, which supports low power usage, but also being able to have a hundred and sixty times more buffering on the chip than they've had in previous versions. And the idea behind this is being able to do a distributed Ethernet fabrics for AI workload trainings across multiple data centers. So if you've got, let's say, three data centers in a geographic region, instead of them all being in the same road and the same rack inside the same physical building, they might be across a campus or something.

Once you go that amount of distance, you're gonna need to have some amount of buffering based on, you know, whether the packets are ready to go out the egress interface and so forth. So having these deeper buffers allows for that. And, course, this would be Rocky v two environments.

The interface speeds here are interesting, Phil. They should be able to go up well, will be able to go up to three point two terabits per second. Now, that doesn't necessarily mean that the Ethernet interfaces and optics and all that stuff are ready to go for three point two terabit ports, but that's what the chip is now capable of. Right? So it future proofs it a little bit to the people who are putting these switches into their environments to be able to do three point two terabit links up to thirty six thousand what they call hyper ports, which are these three point two terabit ports. So just amazing the speeds and feeds that these environments are demanding in order to build these huge distributed clusters that some of the hyper hyperscalers are building.

Mhmm. Mhmm. I love the term hyperports.

Yeah. It's so applicable. You know Yeah. I'm still kinda skeptical on this building a fabric across multiple data centers because one of the things we know about these AI environments is they're very latency sensitive.

Right? Any amount of latency can really impact your job completion time when you're running these jobs. And so I don't know. I guess I'm just still a little skeptical that companies are really going to build clusters that distribute across multiple data centers.

To me, this just sounds seems like the idea of doing a vMotion from data center to data center.

Right? Oh, yeah.

Yes. It can be done. Is it a good idea? Probably not from a design and engineering perspective.

Okay. Alright. I know where you're coming from, and I agree with you, especially if we're considering traditional networking. But we're talking about zero packet loss under heavy congestion, three point two terabit ports, line rate MACsec, the ability to go, I forgot how long, you said many kilometers, whatever.

Mhmm. And I I assume with, you know, all of the the the boxes ticked for latency and for packet loss and all that kind of stuff. And so far as it relates to the networking, not like to the GPUs themselves and that's that kind of thing. You know, the the thing is that we're still in this land this world of brute force AI where we throw more resources, bigger physical data center, more of a footprint with the, you know, the more racks, the more GPUs, and that sort of thing.

And and, yes, I know we talked about small models and and decreasing the size of models and making them industry specific or task specific in order to solve some of this. But generally speaking, the underlying method here is still brute force, more horsepower, more resources. And so you can't, I mean, feasibly build a data center that's under one roof the size of an entire town or something like that. So they're gonna be distributed, and that's that's kinda what's happening now.

Now until we move away from that brute force method and we have more smaller models that that are on par with foundation models and that sort of thing, I actually think that this is this whole, like, long haul interconnect for GPU connectivity is kind of a necessity now. I mean, I I get it. I get it. It it's like it opens up the door to more issues for latency and and things like that, and you mentioned job completion time.

But that's what this Fabric Router is addressing, those kinds of things. Yeah. It's gonna be eye wateringly expensive, of course. And Sure.

But what what's the alternative? It's it is to focus on those small models and doing doing something completely different. I just don't see that really happening. A lot of folks are still focusing on brute force.

I mean, when you say a lot of folks, I mean, I guess we're talking about the hyperscalers here, like Amazon, Microsoft, and Google. Right?

Like, don't see the hyperscalers, you know, yeah, the big AI companies for sure.

You're right.

I guess maybe this just seems like one of those things that maybe those three companies, maybe a couple more, you know, would would do or would need because they're just, like you said, brute force, just building the biggest clusters, you know, tens, twenties, thirties, thousands of GPUs in a single cluster. Right? Just brute forcing it. I just think, you know, you roll down to, like, the the enterprise, Fortune one hundred companies, once they start doing, you know, their own AI fence. A lot of them already are building their own AI fabric. I just don't see them being at that scale. I could see them wanting to be more efficient, trying to figure out more efficient models to be able to fit within the footprint of a single data center versus saying, alright.

Let's just brute force it Yeah.

Throw billions of dollars at this problem and buy Jericho four chips and build it across three data centers in a metro region.

So this just seems like one of those techs that that part of it at least is not gonna trickle down outside of the hyperscalers for some But who cares?

I mean, it's purpose built for those that customer. You know? Whatever. I mean, just like I'm not buying a Ferrari, but there are people that do and they're a very very very tiny small subset of the population and they're gonna buy them.

So there is a market for it. So that's what this is. And maybe we move away from this brute force method where we have not just big data centers, but distributed multiple data centers across a geographic area all doing the same training run. And then that that stops and this this router becomes irrelevant.

Sure. And then Broadcom puts something else out there. But for now, that's where we're going. And so that's that's where I think the significance for this particular router, this Jericho four Ethernet fabric router plays in for sure.

Alright. Moving on. The next article is from Palo Alto's newsroom. Palo Alto Networks buys CyberArk.

CyberArk is the identity security leader, so they're able to do identity management and identity security for a lot of different types of workloads, human workloads, machine workloads, AI workloads. So Palo Alto has acquired them for what would be the equivalent of about twenty five billion dollars. The term of the agreement is that each shareholder will receive forty five dollars in cash per share that they hold, as well as two point two zero zero five shares of Palo Alto's stock themselves. So if you're a CyberArk shareholder and you're you hold on to it through the acquisition, you'll receive some amount of cash and some amount of Palo Alto shares for each share of CyberArk that you hold.

That represents a twenty six percent premium over the ten day average of what CyberArk's stock was trading for as of Friday, July fifth. So last I guess a couple weeks ago as we're recording this. The interesting thing here is I had not heard of CyberArk, though. I don't know if you had.

No. But they're, I guess, a a pretty large company in this space doing identity security and privileged access management PAM. You know, Apollo obviously wants to roll this into their broader security portfolio, you know, for doing a lot of, like, again, identity management and privileged access, allowing people to figure out whether this workload, this AI workload, for example, should be allowed to pass traffic on the network, be able to allow for access based on, you know, what the workload is, things like that. I'm not a security expert, so I don't get all the the bits and pieces of it, but that's my understanding of of the value here.

The customer is being able to integrate this identity management into Palo Alto's broader security portfolio for their things like their firewall and their WAN and some of those kind of capabilities.

Yep. Yep. And networking itself, which, you know, our agentic AI workflows are gonna operate on top of, calling various tools and and interfacing with the LLM brain and and whatever agent system that runs on the network. And so we need to focus less on IP centric controls and more on identity centric enforcement.

And everything flows through APIs now. Right? And whatever CICD pipeline. So you need those just in time credentials and credential authorization, you know, triple a, all that kind of stuff, session control, all of that stuff. Yep. Especially for like our field, if you're using that kind of thing to then also manage your network and your network operations or or IT operations more broadly. So there's certainly that component, not just agents, but everything that we do now that does flow through APIs.

And the concept of the shared telemetry, because you remember you're talking about a huge amount of telemetry that you're gonna gather as far as access control and and all of that and user information, that kind of thing. So there's a lot of information there as well. So you're gonna get you're gonna get kind of more of a consolidation or maybe a simplification of telemetry.

Mhmm.

Obviously, there's a little bit of lock in right there because you're just consolidating it into one platform that Palo Alto has, whether you care about that or not, whatever.

Oh, one person calls vendor lock in, another one calls well integrated solutions, vertically integrated solutions. Right?

So it's, you know Absolutely.

Six one half dozen. I guess it depends on what you're trying to accomplish.

Right? Yeah. And and it does you know, it it it makes it easier, especially if you're a brownfield network to consolidate, you know, network controls and that kind of thing. And if you have, like, a zero trust policy that you're developing, you don't have to, like, forklift upgrade everything. You can sort of integrate these into your current Palo solution.

Alright. Moving on. The next article is from Cisco's newsroom. Cisco keeps taking steps towards quantum networking.

So, Phil, we've talked a couple times on the news podcast about quantum computing and specifically quantum networking. I think as I said before, I'm I'm really interested. I'm really excited to see where quantum networking can take us. Been doing networking for twenty five year.

Latency over a given distance was always just sort of an accepted normal because of the laws of physics. Right? Speed of light and some of those kind of things. You just assume if I'm gonna transmit data from the West Coast of the United States to the East Coast of the United States, there's, call it, twenty five millisecond round trip latency you're gonna have to deal with.

There's just nothing you can do about that because that's the speed of light through glass and that's just, you know, the way it is. Quantum networking, in theory, brings us the ability to transmit data over that distance with zero latency. Right? So Cisco's continuing to do more research.

They have an entire Cisco Labs, Quantum Labs that's started really applying the quantum, as the article talks about, more to the computing side of things. Right? And being able to do things like medical research, drug interaction research, some of those kind of things much faster in, like, a high performance computing using quantum computing. But as part of their research, they're also looking at, like, how can this technology be applied to networking, right, to large WAN networks and so forth.

So, you know, the article doesn't have a lot of details on on new stuff that they've come out with, but it does have some interesting tidbits from Ramana Kompela, who's the head of Cisco Research and a distinguished Cisco fellow, talking about the steps they've taken transferring quantum information to these entanglements. So the idea of entanglement is that there's two photons and if you change the state of one and observe the other one, you can basically know the state of either one. And that's the idea here is if you can apply that to networking, you can actually kind of if you think of, like, we do binary transmission of data.

Right? You can actually figure out whether it should be a zero or a one based on observing the state of it is at least, my limited understanding of of quantum things.

Yeah. I I think it's it's really there's really nothing especially new in this particular article, but it's Cisco bringing it up again. So it's just really interesting to me to see that they're continuing to go down this road and they're building an ecosystem, but they're not just doing that internally. They are working with universities, I think you mentioned.

Mhmm. Standard standard bodies. They wanna ensure interoperability, how to scale this, how to do responsible deployment. And, again, the idea being like, you're not deploying this monolithic quantum computer, but, you know, laying this groundwork for practical quantum networks.

And I love reading the word teleportation in these in these articles. Again, nothing nothing necessarily new new in this headline other than to say that they're still working on it, and they're still expanding this research and and connecting with folks in industry. And, this is something to keep an eye on as we have, but certainly something to keep keep an eye on. And I believe that even if nothing comes of it and it doesn't go anywhere and it dies on the vine, it's certainly important news to keep an eye on how the networking industry is looking at advancing and solving some of the problems that we have today.

Mhmm. Whether it's this or it becomes a different solution, we we don't know.

Yeah. You know, it's interesting. Most of the article talks about quantum networking for the purpose of doing distributed quantum computing.

Right? Okay. True.

Which is one interesting application, and that's clearly what what Cisco's doing the majority of this research for. But to me, the more interesting application, just because this is one more of my interest lies, is being able to apply it to, like, AI workloads. Right? Like, going back to that article we were talking about with the Jericho four SIP, if you had, theoretically a quantum network and you could do Jericho chips and now you can do data centers that are distributed with zero latency, it does make being able to build these huge brute forced distributed AI clusters across data centers you know, even outside of the metro area.

You know, maybe you could even distribute them across the country if you could do zero latency. Again, this is all still future stuff, but that that to me is where this is really interesting. It's, like, not just for quantum computing, but the existing applications that we have on the network. You know, what's the potential impact in how we build and design networks if we don't have to design for latency?

And the impact the latency has on how we build and design the networks opens up a whole different realm of ways that you could straight this out. I mean, you could potentially buy land in the cheapest and coolest place you can buy it and distribute the data from Northern Virginia to some place in North Dakota where power, space, and cooling are really cheap. I mean, there's all kinds of really interesting ideas that this could bring up.

Well, I mean, I'm not an expert in quantum mechanics and quantum computing or anything like that. But if I understand correctly, it's zero latency anywhere in the universe. That's how quantum entanglement works. So it's not like across the country.

It's like another planet. You'd still have the same effect because of the way that particles are are entangled. Spooky spooky, but whatever it's called. I don't remember.

Anyway, I don't know where this is going, but certainly, it's some interesting technology that Be really interesting.

Don't understand in and out. Satellite Internet. Right? High Earth orbit. So you have the reach of high Earth orbit without the latency constraints that that would normally bring. I hadn't even thought about that. That'd be really interesting.

No. Yeah. Absolutely.

So moving on to upcoming events, and we're just gonna cover the next, let's say, next two months here.

We have the Michigan Networking User Group, a part of the USNUA in Detroit on August thirteen.

We have the New York Networking User Group in Buffalo, August twenty one.

And then we have the very first. This is the inaugural California networking user group in San Francisco on August twenty one. I'll be attending that, and I'll be speaking about some practical uses and implementations of AI, specifically in network operations. That's what I'll be speaking about.

If you're in that area, you can register at the US anyways website and just go to the events page or scroll all the way to the bottom. You'll see it. Then we have the Ohio networking user group in Cincinnati on September fourth. Justin, I believe you're gonna be there speaking?

Yep. I'm doing the keynote for that one on content that broke the Internet. So one of the things I find kind of a nerdy passion of mine is being able to combine current events and media things that are going on and what we what we actually can see about it on the Internet now that all this content is, of course, distributed out to users over the Internet.

So Yep.

Great.

We also have Yada Las Vegas, September eight through ten. Kentik will be there, and I believe Steve Muse will be speaking on behalf of Kentik. So that's that'll be good. And then we have DCD London, September sixteen through seventeen, which is data center dynamics. A lot of stuff about, you know, AI workloads in the data center, but also in data center infrastructure and compute and all of that other stuff. So lots of great events coming up. We will start seeing that list expand, especially as we get into the fall and the conference season really picks up some steam.

So for now, thanks so much for listening. Those are the headlines.