Episode 410 – Why Data Movement Defines Hybrid & AI Success – Why Is The Internet Broken?

Ep 410 - Why Data Movement Defines Hybrid & AI Success
=======================================================
Justin Parisi: I'm here in the basement of my house and with me today I have Glen Deckheyser. So Glen what do you do and how do I reach you? 
Glenn Dekheyser: Well, I'm in a basement of my house too.
Justin Parisi: Are you? 
Glenn Dekheyser: Yeah, I am. I, it's is it 
Justin Parisi: snowing where you are? 
Glenn Dekheyser: It's cold. I don't know about snowing. It's cold here too. It's definitely cold. 
Justin Parisi: School will close for us because we're wimps. 
Glenn Dekheyser: They opened us to schools longer when it's cold and snowy up here. So LinkedIn's where I'm doing most of my social stuff today.
I am on Blue Sky. But LinkedIn's where I do most of my industry stuff. 
Justin Parisi: Okay. Excellent. So you work at a company called Equinix and we've worked with them in the past with cloud partnerships and that sort of thing. So what exactly does Equinix do?
Glenn Dekheyser: Our tagline's the world's digital infrastructure company, right?
We've got 276 data centers globally. All connected via Equinix fabric, which is a private digital interconnect. And we're kind of known not just for that, but also for our ability to connect customers to the clouds and resources that they need very, very quickly through Equinix fabric or even through classic fiber cross Connects. Equinix houses some of the world's most critical ecosystems. You know, financial ecosystems, healthcare content delivery, hyperscale clouds. And just to give you guys an idea of the scale of where we sit we like to say that 95% of all internet traffic around the world at some point in life will traverse through an Equinix data center.
So people don't realize how much they use us without knowing us. I like to say that if you pick up your iPhone and you turn it on, you open up an app. You've probably used this 10 times and didn't realize it. But we also happened to have a large amount of the cloud on-ramps. We were the first to have a cloud on-ramp through a direct connection with AWS. And NetApp and Equinix got started with that first relationship you were just alluding to. I think it was back 2015 with NetApp private storage. So it was the idea I could put a physical NetApp close to the cloud connected over this link, which at the time was maximum one gig a second. You could do over one gig a second, but they were doing it, to do iscsi connections, to run virtual machines in the cloud and run them with the data stores back over one gig. Like I said, it worked for the things it worked for. It was a great solution that was a little bit before its time. Consider now that we've got a hundred gig connections to the cloud in local metros in some cases. You can make more by going direct. But the clouds typically put their switch infrastructure at Equinix, to get close to the customers they want consuming. So a customer that has Equinix as their co-location, has their compute, has their storage, has their IT core running at Equinix can connect to many clouds around the world more than any other provider by a long shot. I'd say by a factor of three or four. I don't know what it's these days, but you can connect there with very low latency, as low as you can get. Sometimes getting ping latencies from your physical presence to the virtual machine in the cloud, lower than it would be from the virtual machine to the virtual storage in the same service. So you can get somewhat parity on latency. And then you just gotta worry about the throughput and you can get a lot of really cool use cases, especially with AI and moving data around.
We will get to that, but you start to see why NetApp would be very interested in harnessing that power given NetApp's supremacy in data motion, right? For a very long time. It's one of the things that I always exploited NetApp at great advantage for my customers was getting data around in consistent ways that it could be used, read only, ReadWrite, cloning, all the cool things you could do and doing it efficiently with deduped and compressed SnapMirror or vaulting, or now flex caching has really become a big deal. So the possibilities are endless of what you could do with this. But to give you just another idea of the scope Equinix has 12 metros around the world that have six public clouds adjacent to that metro, and the six being top three that everybody thinks about, you know, AWS, Azure, Google.
But then you also have IBM Cloud, you have also Oracle Cloud, which is really showing some amazing things lately. And Alibaba. The rest of the world has zero of those, right? There are no other provider that has all six in a latent way. So once you start talking about multi-cloud beyond hybrid cloud then Equinix becomes almost an imperative partner because you don't know where you're gonna connect tomorrow.
And so you wanna make sure that where you're gonna place your infrastructure core. That's not easy to move, especially if you're moving data around that it has all the places to go. So that's kind a perspective from a person who comes from the storage industry Equinix's value is being able to create a data hub that'll hold what I'll call the authoritative copy of that data.
But you call it what you wish. And moving it from cloud to cloud. So if you wanna have a data set in AWS, that's great, but you know, you get a petabyte in AWS and you wanna move it out. That's expensive. Time consuming. If you had a petabyte on-prem, and if you couldn't use it for whatever reason, over a cloud over it in a hybrid way, NetApp can put up fSxN. Up in AWS or, use ANF or use Google Cloud Volumes and project that data from on-prem to there to be used and use it as long as you want. Updates cascade up into the cloud. You're using your data up there, and when you're done and you wanna move it from one cloud to another, you wanna fire this cloud and go to that cloud.
Or you just want to use a service in a different cloud that this cloud doesn't have. All I gotta do is delete it, right? Mm-hmm. No egress, no time. I'm in control. I have all the leverage over my data and I'm not subject to the price hikes where all of a sudden if I've got my only copies up in the cloud, or I've got two copies up in the cloud and they're killing me on transit costs for replicating and they decide they're gonna raise my costs, I'm kind of screwed because it's gonna take me a long time to get out.
It's gonna cost me a lot from an egress perspective. There's ways to get around this and NetApp has all the tools that customers really need to bring that to bear. 
Justin Parisi: Basically, what I'm hearing is there's this idea, this goal of a cloud utopia where you don't even think about it being a cloud.
It's just, I use it, I consume it, and I don't care which cloud it's in. I don't care what storage is there. I don't care which computer is there. It just works, right? And that's kind of where the data movement stuff comes into play because that's what we can do on the back end so that you're not thinking about where the data lives and what sort of round trip time I need to worry about and how many hops are there and what's the network bandwidth and all that extra stuff that maybe you don't want to deal with so much. That's kind of where you want your cloud to end up and that's what Equinix provides, right?
Glenn Dekheyser: The message that we try to get the customers is that you need to know, you do need to think about all that stuff. And the folks I talk to mostly, in the world of storage have typically not thought about that. And if you're working with a cloud provider, they're not gonna point you to that because it's not in their best interests to get you to think about that. But it is very, very important that if you're talking hybrid multi-cloud, or even just hybrid cloud, it's extremely important to understand the concepts underpinning how data is getting from the network interface of your on-prem to the physical and virtual network interface of the cloud version of ONTAP, pick your flavor.
And that really kind of alludes to how I got to Equinix in the first place. So, back when I was in the partner world, one of my last projects before I left, and you'll understand why in a minute. I was working with a large pharmaceutical company, and I was working with an HPC group and they're doing DNA folding and all that kind of stuff. They're sequencing DNA, sending that data to different places, and they wanted to take their DNA sequencers and take that data that they gather on-prem, send it up to Amazon, and use a DNA sequencing application up in the cloud in infrastructure in EC2. Many instances of it against this storage. They wanted to get rid of their compute on-prem. Perfectly valid use case happens all the time now. And I was doing this project, it should have been a slam dunk. I mean, the data movement, we understood that.
We understood what the data was. We understood where it was going. And it just couldn't work. It should have worked. Everything was set up. It looked like it was set up right. We were going over the connections from the customer's premise to AWS. They were using private interconnection. The HPC group wasn't the network group. We didn't have the network group involved, probably should have. And they just weren't working. So it turned out that the connectivity between the compute, that whole VPC, and our on-prem, when you did a trace route on it, it was taking upwards of 300 milliseconds of latency.
Justin Parisi: Is that bad? 
Glenn Dekheyser: It is not good. It's not good at all. When you're at Equinix and you have your storage in the same region, cloud adjacent, you're typically talking, let's just say less than five, but usually less than two milliseconds. So you're talking about 300, that stuff getting ugly. But, the Amazon was showing all green lights and the NetApp stuff was all showing green lights, but nothing was working. The customer had no idea how their network was set up. The network group, when we finally brought 'em in, really didn't have much of an idea of how cloud networking worked and how it was routing, even inside their own network.
And then getting out to Amazon, they didn't understand where they were connecting to from Amazon. And to make things worse, it was going cross region twice in Amazon, which created an enormous cost. They got a bill that they didn't know where it was coming from or why. Because Ingress is free in Amazon. So they got this inter regional bill that was just totally unbudgeted. The customer was not very happy with this. And I realized I was missing a dimension of architecture in this, and this is like six years ago, seven years ago at this point. Talking to other folks active in the NetApp community, I realized most people didn't think about this. They said they would've run into the same problem, 'cause they just figured the customer guys has this figured out. And I will say that if you're a storage guy and you're doing anything hybrid cloud and you expect your customer to have optimized or understand even their hybrid cloud networking, you've made a big mistake. Assume they don't, and when they don't, obviously your stuff is probably not gonna work the way you think it's gonna work.
Your customer may get a bill or they may not just get the performance they want. Things just aren't gonna, come together. You need to be able to put your customer in a position where they can succeed not just with this project, but with the future projects and keep them flexible.
So I think the time is now for storage people to make sure they understand cloud private connectivity and things that have driven that reason to learn, especially now as the world you'll see it more outside the US but certainly you see it here as well.
Multinational corporations, sovereignty has become a huge topic. Customers are really increasingly looking to get that data on equipment they control, in locations they can access so that they can claim the sovereignty of it. I'm not saying they're looking to extricate the clouds necessarily. They are certainly looking to make sure that if they have to get out, it's easier than it has been in the past. You start to see a greater interest from some of the public clouds of interconnectivity between themselves so they can move, in some of the more on-prem solutions like Azure Local Outpost, you see more interest in your Google, DC GDC, you start to see more interest in those tools. They've even become more amenable to allowing you to have your own storage that will connect to these things on-prem in your site. They never wanted any of that. It was either all them or go away. Right. All of a sudden they're playing a lot more nicely with customer on-prem stuff. This is all great. But from a storage guy, it's not as simple as it used to be from a connectivity perspective.
It's not as easy as LAN or backend storage networks. That was all we really used to care about. Now that data's getting moved out to all these different hybrid clouds and now you add the neo clouds that you're also gonna be sending data to perhaps for inference, for agentic, pre or post training, all that stuff.
So you need to understand how this private interconnectivity works. This is where Equinix lives. We are the center of that world that has the most possibilities to kick to all of these things. Neo clouds, public clouds. And that's kind of how I got to Equinix, was I realized by missing that dimension. I happened to have a meeting with Equinix at a customer the week after that, that experience and heard their Equinix fabric pitch. And I'm like, I see something here. This is what I think I've been missing. Dug into it a bit and really went all out and said, please hire me. 
Justin Parisi: So, when I said earlier about this cloud utopia and people not caring, it was more about the personas, right?
So your persona of your end user is the one who shouldn't care. The persona of the storage, admin, or the network admin, of course, is gonna have to care somewhat. In your specific scenario, your thing that you worked on, what ended up being the fix?
Glenn Dekheyser: If you, I mean, it got fixed after I left, but Oh, it, it, they ba they essentially came to Equinix.
Honestly, they weren't working. They did eventually come to Equinix. I didn't have anything to do with that project, but, I did see it. They ended up rationalizing their network and getting low latent and it ended up getting architected the right way underneath the stuff that we had originally put in. So the stuff we sold didn't change. It was just the direct connect configuration going outta the right region, making sure that the last mile stuff was set up correctly, their internal routing was set correctly.
Stuff that we say it's the network guy. Well, that network guy's gotten a lot more important to us Yeah. Than they ever used to be. Yeah. So better make friends with them. Stop blaming 'em for stuff. 
Justin Parisi: So what sort of best practices or tips and tricks are you recommending for networks when they're trying to interact with either private or public cloud? What are the best things that people can think about before even starting to get there? 
Glenn Dekheyser: I wish there was a universal answer for that. But what we do at Equinix, when we have a customer that wants to optimize that outcome, we do a workshop with them and it's a free workshop, but we spend a day with the customer sometimes two, going through their use cases, going through their current configuration, where they're at, like even that end user you were talking about before who shouldn't have to know about the storage. They do need to know where their data is, discovering where the data sets are. And we'll get into AI in a minute. 'cause I think AI really changes the game on all this. But let's talk legacy it, standard work cases where these applications are, where their users are, where their customers and partners are. Anybody they connect to digitally. All of this stuff is part of the workshop. And then you figure out, okay, well how are you doing things today? Are you using NPLS? Are you using legacy network stuff? We've seen customers save 80% just by replacing NPLS with Equinix Fabric as their WAN backbone. So there's easy wins that can fund other projects, but more than that, it's about, okay, well I wanna use SD wan, right? And again, I'm talking SD WAN on a NetApp podcast. I mean, right. Then storage guys are like, I don't wanna talk about that.
But it is important to understand how your users are gonna get to the data that you're provisioning. And also, how that data's gonna move to the different sites for dr, for batch reporting. How we're gonna use all that NetApp goodness to affect these business outcomes.
That can't be done with other platforms. That's really key. NetApp can do things other people can't. But it doesn't do it by itself. It requires knowledge of how to exploit those features on the compute side, but also how to manage that and make sure it's working all on the network side. It does take a team of folks to come in there and help you work through all that, who have experience with it, but then lay out that plan. And most of the stuff that has to get done isn't rocket science stuff. It just knowing what has to get done, most people just don't know.
And there's no book you can go buy to go figure that all out. So I would just encourage to make sure your partner is Equinix aware that they can bring our folks in to help do that workshop that works them through, okay, here's where you are today based on the business requirements and the things you need.
We sit down and do all that, and we give them here's what it would look like. Here's your brain here, here's your brain on NetApp, and here's your brain on Equinix. This is what your infrastructure would look like if you redeployed the network with us, here's how much it would cost. Here's how much you would save or not, maybe but here's what you'd be able to do. And usually that's when we also start the AI conversation. Well, where are you going with AI? And answers inevitably, we have no idea. Most of these customers are still looking for the answers.
Yeah. And that's really where I think Equinix and NetApp together can shine. You're not supposed to know the answers, but when you don't know the answers, you wanna put yourself in a position to be able to experiment, iterate, and then deploy. That's the strength I think of the two names put together.
NetApp being able to go and take data sets consistently, move them to the three clouds and first party offers, which I think is really, really good. Either doing it in a caching way or using SnapMirror, which is now one of the new things that came out that was at Insight.
I was really excited about that. Being able to, in the core, have an AI data platform to exploit with A IDE and give your data scientists a place to work on, then you could then take that data and embeddings and stuff and move them to different places. Allowing you to prove out value, do it in the cloud, maybe you gotta go buy your own GPUs.
And let's talk about the edge as well. I've got 74 metros that those 276 data centers are sitting in. So I've got an edge that instead of going to the far edge and go into an office, I can go deploy it at a Metro edge at any of those places. Again, connecting over fabric, I can take this data that's in this core, that's on an AFX cluster, let's say, now curated through AIDE and NetApp can go and reliably, confidently push that data to any of the clouds, the neo clouds, or any of these edges over Equinix fabric. And by the way, all of that can be configured through Terraform, Ansible. It's all API driven. But, you're in New York, you want to go to Seoul and start moving data over there. We can have that done for you in five minutes. It's really that simple. And it will always be the least latent connection you could possibly have because it's on our own private network and they're gonna be long haul fiber connections that aren't going through a million hops. It's not like you're going through a big NPLS network or over the internet. Right. Yeah. Most importantly, you wanna talk about sovereignty. Today we offer this in the us, Canada, and Japan, but we're coming up on other regions. There's a lot of reasons you don't want network packets going over national border. I actually just earlier today finished our interconnected podcast on post quantum encryption. People are collecting data that they know they can't read today, but they know once they get the algorithms, they're gonna be able to read 'em, eventually. So right now, certain organizations are collecting on this data. 
Justin Parisi: They're data hoarders 
Glenn Dekheyser: basically. And so one day they'll be able to go and decrypt that and learn certain things most about authentication stuff.
But, Canada doesn't want financial data dipping into Patriot Act territory, and the United States similarly doesn't want data going up into Canada for certain type. We have the capability of saying on our fabric, Hey, if you're in Vancouver, you're going to Toronto.
We can only go through our Canadian fibers. We can do that. We're like the only ones who can do that. So it's these kind of sovereign things, when you start working with data and you need to ensure certain things. You may not need that today, by the way, but know that one day you might be asked to do that.
The capabilities are here already, so it's all about the possibilities and the unknown. If you ask any enterprise what's our AI environment gonna look like in three years? Any answer they're gonna give you is wrong, first of all, if they give you one.
So giving them the ability to experiment, iterate, and then operate. I think it's priceless almost, giving them that ability. NetApp provides that functionality from a data perspective, that can leverage our platform of interconnection and locality that can the two of them together, create that value for an enterprise to go and exploit now, and gives them that optionality. The way I got customers to appreciate the NetApp platform who weren't aware of it, was that, look I give you with NetApp the privilege to change your mind and not change everything, right?
You can literally go from block to file and not change a thing. You wanna do block with this data set and you wanna go NAS with it later. Great, no problem. You wanna go change from NFS to CIFS? Great. You wanna do both Great. And all these things you can do. NetApp is really the first to the party with that kind of stuff. Put that on top of a platform like ours that allows you to do it globally in a distributed manner and it lets you affect what we call distributed ai. And I really do believe that it starts with the data, right? GPUs can use it wherever you want, but between NetApp and Equinix, customers are able to create, I'll call it a composable AI factory, where you can pick your GPU from the platform of choice that makes sense for your economics and your performance, and either move data up or use it in place. You have all these different options of doing it. When conditions change or when economics change, or you have a new business need, you can go do it really quickly, and you can repeat these AI outcomes. I can't think of a combination of products and platform that can do that more than NetApp and Equinix can. Yeah. While NPS back in 2015 had the right idea, it was way too early. We're gonna be coming back to a cloud adjacent AI platform with NetApp and the partners.
And we are starting to work with NetApp on some of those things too. To get those to market with our common partner set. As a matter of fact, we happen to have a pretty overlapped partner community, so 
Justin Parisi: yeah. Yeah. The AI in the cloud thing is interesting because you're dealing with these GPUs, which can get very expensive, very quickly and they may not be things that you're using all the time, so it doesn't make a ton of sense to buy them. 
Glenn Dekheyser: Correct. 
Justin Parisi: So then you start thinking about your lease model, right? Like leasing a car or leasing a brand new BMW and then every three years you get a brand new BMW and it's even better than the last one.
And that's kind of the way you would handle A GPU here, because if you're not gonna keep it, right, right. That's the point. So that's where you get your cloud burst use cases. When you need to increase your production somehow, you need to add extra GPUs. Maybe there's not GPUs available because honestly the clouds are buying 'em all.
Right. Wow. Does Equinix provide it's own kind of neo cloud solution where it's creating its own GPU as a service? 
Glenn Dekheyser: No, we're not. Neo clouds are ecosystem that customers can connect to. We do have some neo clouds that locate their GPUs at Equinix.
Some of them do. Okay. You might have seen some great announcements from say like groqcloud, that's with a q that groq cloud. So they have their own custom chips that they all make, and they're focused mostly on inference. We've seen some customers have some great success using those guys and reducing their cost from cloud tremendously. But that's part of what I'm talking about. You were doing it up here and now I wanna do it over here. If I put my data up there, it makes it a lot harder for me to get my data out here. It's gonna cost me money to do that.
So there's a barrier to business there. I shouldn't have that barrier if I had my core here on real storage in the middle or even adjacent to say, a region that had multiple neo clouds and a and a cloud. All of a sudden my bets are all on I could do everything at the same time.
Use cases one through five. I could put in say, groq cloud. Use cases five through 10 I can go use my hyperscaler 'cause it makes more sense. But I don't have to choose all or nothing with these things. As enterprises get to a hundred use cases, every one of those is going to probably consume a different combination of resources, especially when you start adding on the agentic layer onto all this. So how am I connecting? What am I connecting to with agentic? Well, sometimes I'm connecting to my own stuff, which might be in my legacy data centers. Okay, great. That means my GPUs as a service provider has to connect to internal legacy environment, even though I didn't have that necessarily, it might have to connect to ServiceNow, Salesforce, some of my SaaS providers. And sometimes in the industry specific world, I might have in the oil and gas world, I might connect to some IOT provider or a C3 or one of those guys. I have to connect to all these things. How am I connecting to all this stuff? Well, that's what we do. At Equinix, you are going to have the greatest probability of success for all of your different use cases if you're working with us. We have customers that only put their networks with us and don't even have their data centers with us. They connect their data centers into us, and then from there they can break out and do whatever they want.
All are welcome, right? That's all great. So utilizing us for what we're best at, and especially with struggling with ai. We connect to some of the larger GPUs of service vendors that don't house their GPUs with us and for good reason. You get these data centers, the one gigawatt data centers, typically those aren't gonna be with us typically 'cause we're more of a retail, we're like Grand Central Station for AI or it, right. You can have a storefront at Grand Central. It's probably gonna cost a little bit more than usual, but think of the foot traffic you're gonna get. So if you want that foot traffic and you want the people that are coming into that train, you wanna go locate yourself right next to the train to on Con or whatever. That's Penn Station, not Grand Central Station. Don't kill me New Yorkers. But the point is you want to get your service closest to the people who are gonna consume it. That's just basic stuff. So we're kinda like Grand Central Station, Penn Station, union Station.
Pick your train station for the world of it. And that's why we had 95% of the internet traffic passing through us at one point or another. 'cause even when it comes in from the internet from somewhere else, it might go privately to one of these other companies that you're working with.
So, working with us is just gonna give you a greater chance of success of whatever future projects you're gonna do. In ai, I just think it's gonna make a lot more sense. We have a lot of the things at Equinix that help you compose these solutions together. I'll give you another example since it's storage related, and one of my roles at Equinix is I oversee a part of our business development throughout the entire storage ecosystem.
We already have some customers signed up on this, but essentially it's tape as a service and served out through an S3 interface. It's not glacier competitive, it's different single tenant tape where you get your own tapes, but we're not mixing data in with other customers, but it's 2025 and I'm actually really excited about tape.
Kind of weird. Right? Weird. So, yeah. Well, you know, circumstances, supply chains and stuff have made it so that all of a sudden these shingle drives gonna be a little tough to get for the next couple years. The hard drives. So all of a sudden you gotta start getting data onto those drives for your AI endeavor.
You gotta get that old, stale data out. Yeah. We also believe that AI is going to drive a lot of tape requirements, checkpoints, raw data sets, lineage, things like that. And so you're gonna have to store these things for many reasons. Not just eDiscovery, but a whole lot of other reasons.
So we think there's gonna be a big comeback on that. Well NetApp works in that world, and they have the whole ecosystem. Now you've got your AI data pipeline, which could have, let's say an AFX cluster and AIDE running, right? Well, I wanna manage that data. Let's say I use a Commvault, or I use Veeam or another NetApp partner to go back that stuff up and get it to tape for long-term AI archiving. I can bring it back and automate it back with my AI pipeline. Who wants to run a tape library anymore? Nobody, right? But you still need to use it. So what we're trying to do is put together these solutions. I hesitate to call them managed services, because our customers for our components will actually be managed service providers who don't wanna manage tape libraries physically, but we can do that for them, right? We are managing the private AI stuff where we'll manage up to as much of that a service provider doesn't wanna do.
For instance, managing up to base command or mission control in an NVIDIA cluster. We'll take all that and manage all that, including if there's liquid cooling involved. Which we haven't discussed at all, and Equinix is ability to bring liquid cooling into this world with the new Grace Blackwell stuff and beyond and the crazy requirements of Vera Ruben in the future. But as we start getting to these high power stuff, equinix is ready. But most folks don't know how to manage a liquid cooled environment. They just don't. And even a lot of the partners don't know how to do that. And why should they, how often do you get to do that? 
Justin Parisi: Well, that's the thing. You set it up once and you're not probably gonna set it up a lot after that. So why? 
Glenn Dekheyser: Well there is a lot of work and maintenance and operationals.
Justin Parisi: Yeah, there's maintenance for sure. Right. So you could either hire an entire team that specializes in that, for probably 10x the cost, or hire somebody who already has the expertise, they already have the knowledge and they do it for you. And that's I think the draw of this type of stuff is. Licensing the expertise is really what it is.
Glenn Dekheyser: Yeah, from a liquid cooling perspective, I think that's one of the things that's gonna drive a lot of the neo cloud success, and I like to put this in terms of if you're a storage guy, you'll get this. We've dealt with a lot of it with NetApp in the SnapMirror world, both sync SnapMirror and Async, metro cluster. I've had conversations with CIOs who say, I want zero RTO zero RPO. I want it automate it too. As soon as a smell a disaster, I want it failing over. I'm like, okay. You want me to propose that solution for you? You sure about that? And you want the solution? Here's what we're gonna do. I'm gonna charge you $25,000 to make that proposal for you. If you go for the proposal that $25,000 will be credited.
If you don't, you owe me $25,000. 'cause I guarantee you're not gonna wanna do this. And they never do, because the conversation always goes, they get the proposal for the zero zero and they go, oh, maybe I could deal with two hours downtime. Or maybe I could deal with 15 minutes of lost data on some of those data sets.
I'm like, Ah-huh. Right. And I think the same thing with liquid. You gonna get some of the same thing with the AI liquid cooling. So you think, you want a super pod of of Grace Blackwell. Okay, maybe you need it. There's the price of the super pod, but then there's also the price of the liquid cooling to bring that to bear, which is not zero by any stretch. It does increase a bit and people aren't ready for that. They're also not ready for the operations of that and the ongoing cost of that. So I think some of that's gonna drive new cloud, but I think you'll see Equinix provide solutions that will accommodate customers that do want this.
And like I said, we'll have the solutions help them manage those things they don't want to manage, or the service providers or partners to manage and take those things off their plate that they don't wanna be doing. Mostly the physical stuff, the physical infrastructure stuff. And so I think that's one of the things you're gonna start seeing is you're gonna start seeing people who never had an appreciation for the physical side of this world that we live in start to have a little bit more appreciation for it because it matters. We never used to talk about power as much as we do today. No one ever talked about water flow like they do today. Both of those things are top of mind when you're talking about AI now. Even if you're only doing inference, so these are things I never talked about before I came to Equinix, and even in the first year or two, I was with Equinix I didn't talk about it much because AI wasn't really being pushed yet. Now all of a sudden, it's all anybody wants to talk about. It's amazing how all that's changed and it would be a really good idea for the storage practitioner to become somewhat fluent in these conversations because being the first step in that journey to AI puts you in a position where you're gonna need to talk about this stuff and paint the picture of what this is gonna look like going forward if you do things this way versus that way. It's important for that storage guy. You're in a driver's seat for the entire enterprise journey. If you don't realize it, you are. And so be ready for that conversation and make sure you're enabled on that. 
Justin Parisi: I would imagine that conversation's very similar to whatever a crypto miner would have to talk about.
Right? Because you're dealing with the same kind of sheer processing, trying to cool down this inferno of processing power with whatever you have, right? You can't really buy a misting tent for it. It's not gonna work. 
Glenn Dekheyser: A lot of people want the submersible stuff.
Right. Or the iers, 
Justin Parisi: the Microsoft actually did that with, I guess, some of their cloud data centers. They just basically stuck it in the ocean. 
Glenn Dekheyser: Yeah. And then they pulled 'em up. They said, now that was a nice experiment. 
Justin Parisi: Next we're done, we're done. So with agentic ai, away from like the whole liquid cooling thing, but it got me thinking about data sovereignty.
I would imagine there's gonna be some challenges with that, and I would imagine they're gonna be very similar with the challenges that we saw with iot. So talk to me about how Equinix might address some of those challenges. If there's even challenges. Maybe I'm completely off base. 
Glenn Dekheyser: Well, there's risks that need to be managed, and this is where private interconnectivity versus internet matters. Mm-hmm. Again, it comes down to where those packets fly. If you're in a regulated industry, you really don't want that stuff going over the internet. And a lot of people when SD WAN first came out said, Hey, I can do a private network over the internet. The problem is that once a packet leaves your firewall or router, depending on what your egress point is, you lose all determination of performance. You lose determination of path, of uptime. None of that stuff between your two sites is under your control.
So some branch stuff might work that way. It depends on what your own internal SLAs might be. If you're a regulated industry, that's not gonna cut it. Private interconnectivity really matters and that's why a lot of companies stuck with NPLS, where there was a network provider watching every link. And using Equinix Fabric, you get that same watch only with a simpler, less latent, more deterministic network from a performance perspective, a lot fewer hops and amazing functionality in there. So we give the customer the ability from an agentic perspective, and we're building I should say we've already kind of built it, but it's kind of in its version one, of an AI inference network that has agentic built into it and allows for intelligent use. I should, I'm trying to figure out the best way to say this, that I can say it. 
Justin Parisi: Is it basically a traffic cop, like an AI traffic cop? 
Glenn Dekheyser: Yeah, there's a lot of utilities out there and tools that allow you to choose your foundation model being run as a service. But none of them manage the interconnection of it. It's all goes over the internet. So imagine if you could pair that with automated private interconnectivity to those same resources that you'd pick, as a hypothetical example, that would let you go to Core Weave or Nebeus, but imagine the ability to go and have that privately without having to do that manually every time, which is annoying and being able to do that via intent-based agentic workflow through MCP. Like saying in some prompt, I wanna connect to AWS and Azure outta this region, and I want to connect to my locations at Equinix, in Dallas, and in Silicon Valley. Type it in and it's done. Imagine that, right? Yeah. Because right now, you wanna go get this done, part of the problem is that does a storage guy really need to know BGP or does he want to know BGP to make all this stuff I'm talking about before work? That's been kind of the challenge, is that okay well oh, so you wanna do what used to be Blue XP in a NetApp console, right? You wanna have your NetApp send this volume a copy to this VCP this FSxN in this region, and I wanted to send it to this ANF in this region. Okay, sounds easy enough. But in order to get that done, I've gotta go fire up the FSxN.
I've gotta go fire up the ANF and I've gotta do all the security stuff. Then I gotta go fire up the direct connect. I gotta go to Equinix Fabric, I gotta get the BGPL set up. Then I gotta go to my internal network. Get you start to see the, starts to get, like I'm a storage guy. 
Justin Parisi: Yeah. 
Glenn Dekheyser: I don't wanna do any of that.
Right? Small layers. 
Justin Parisi: It's a lot of steps and it's easy to miss steps. It's easy to make mistakes during that process.
Glenn Dekheyser: And I don't wanna talk to the network guy. I just wanna make, I just want it to work! 
Justin Parisi: They're jerks. I mean, they're mean to me. They always tell me I'm stupid. 
Glenn Dekheyser: I thought that was the DNS guy. Anyway, but imagine being able to type all that in a prompt, I want to do this, I wanna make this volume over here, and I want this to do like this.
And I put it in a prompt and it's all done. Right? So that is where it's all going. And if anybody from any of the partners listening wants to learn more about that, they can reach out to me and I can help 'em set up with some of our with our folks who are building all that.
Justin Parisi: Awesome. So another thing you've been involved in is the NetApp A team. Yeah. So long time. Yeah. So tell me a little bit more about your involvement with that and what sort of things that they offer to people who might be interested either becoming a part of it or engaging with them.
Glenn Dekheyser: Well, if you're someone who knows a lot about NetApp and bringing value to customers using NetApp's unique value proposition and can really, demonstrate that, not just in words, but, actually you're a practitioner and you can show it and you're experiencing on the ground what NetApp is bringing to the table, all the awesome things and the once in a while you see things that don't work. Right. Right. If you're interested in, having an impact in that community where there's a lot of other people just like you who are interested in optimizing these customer outcomes with this technology, that's what the, A team is there for, right? The A team is there to give you that vehicle to first of all get the information from the people who are creating that technology, being in the room so to speak. You're not actually in the room.
You're in the room next to the room. But it's close enough and you actually can have an impact if you have experiences and you need to talk to this specific person to have direct impact into fixing that, or get input or give them input, the, A team is a great vehicle for that.
Now the, A team is also a great vehicle for your own personal brand management. If you're a blogger, if you do a lot of tweeting or whatever that's called these days posting, I guess x-ing.
Next thing, whatever crossover or I do just LinkedIn posting is what I do now. 
Justin Parisi: Blue skying. Do you blue sky? You don't even blue that anymore. 
Glenn Dekheyser: They had a different word for it. I wasn't happy with the word, so I didn't use it. But also to the internal impact if you wanna have an external impact and you can get that stuff out there.
And what the NetApp A-Team does, is amplify your presence online to make sure that the people know that you're an expert. And it really helped, it helped my career tremendously. 
Justin Parisi: Yeah, it does help careers. I mean, we've had people get hired to NetApp from there. If you're going somewhere that deals with NetApp, having that kind of credential where you're a trusted NetApp advisor.
Yeah. You know, I mean, that's money in the bank really. I think it's honestly in some ways better than having a certification. 'cause certifications are a dime a dozen. It helps, it definitely helps, but have having a community, people that you can bounce ideas off of and that sort of thing, I think really is important.
Glenn Dekheyser: Yeah. It's not a huge community. It's one of those things where you really have to be someone who understands not just what NetApp does, but almost why they do it, why they made these choices and what's the value proposition to the customer for having this particular feature set or doing things this way versus this way. More than just, oh yeah, I'm an NCDA. It's more than that. 'cause anybody can pass a test. Well, not anybody, but the point is that this is for, if you're passionate about it and that's really the key. All the folks who've been on the NetApp, a team over the years have all had a passion for what they do.
And some of those meetings, especially with product, some of 'em got heated because the people who were the A team folks would get pretty, I don't wanna say angry, but when they were talking to some product folks, especially when Flash first came out, there was certainly some heated conversations, not because anybody wanted to yell at somebody, but customers need things and, you have the ability to get that to them. And so these people are really passionate about customer outcomes in general. So it's a good bunch of folks.
Justin Parisi: Yeah, I was gonna say, it isn't all rah rah, rah. I mean, we have some very grumpy a team members. 
Glenn Dekheyser: They're No, I wouldn't say they're grumpy. No, they're grum and they can be grumpy. Well, heck, I've been grumpy. But it's all grumpy towards the right goal. 
Justin Parisi: Yeah. Yeah. It is grumpy for the right reasons.
It's like, yeah. Anyway. All right, cool. Did you have anything else you wanna cover before we close it up? 
Glenn Dekheyser: Just one last thing. I am the host of Equinix's Interconnected podcast. This is not a sales podcast. It's a really cool podcast where we go and talk about a specific emerging technology. We get somebody really cool to talk from the industry or academia about it. And we don't talk much if at all, about Equinix. It's mostly just about that technology. I got to interview a Nobel Prize winner in quantum computing a couple weeks ago, which was kind of cool.
That's pretty awesome. No, it's fun. And it's well produced and it's a follow on to our Trace Route podcast we did a couple years back, which was award-winning. If you haven't sampled Trace Route I encourage you to do so because it will give you a lot of background, especially if you're a storage guy, that the Trace Route podcast, which is all done already, but it's all still valid information goes through the history of internet, cell phone, 5G, the whole thing, and infrastructure about all those things that you probably don't know about.
It was really meant to talk to the younger generation coming up. A lot of these kids never even seen a server, they're doing coding. About a data server, server. It's all server. Now everything's serverless. Right. So, you know, to, to give them kinda that background, Hey, this stuff doesn't just work. There's a lot that goes into this. And every time I take somebody on a tour of a data center, they're like, whoa, I never realized it was like this. And then this one's more about the emerging technologies and their impacts. And so it's been a lot of fun to do. I'm not the only host, but I've done about four or five episodes now and I think there's like six or seven.
So there'll be more coming out this month, but it's the Equinix Interconnected Podcast, so check that one out. And that's been my other project. 
Justin Parisi: We'll add a link. So that people can get to it. But again, Glenn, thanks so much for joining us. Thanks for talking to us about interconnectivity.
That's another one of those things that the young people probably don't think about. Right. 'cause everything's mobile now. Everything's wireless. 
Glenn Dekheyser: And everyone's just in their house and playing by themselves on their computers now. Right. Yeah. No interconnectivity whatsoever.
Justin Parisi: Well, I mean, that's a different kind of interconnectivity. Yeah. Anyway. All right. Well, yeah, again, thanks for joining us and if you wanna reach out to Glen, he's given the contact information. We'll include it in the blog as well. But yeah. Awesome. Thank you. 
Glenn Dekheyser: Appreciate it. Thanks for having me, Justin.