Public Cloud Repatriation - Is it real?

Machine-generated from the episode audio and not hand-corrected, so names and technical terms may be imperfect. The audio is authoritative.

Welcome to episode 115 of the CTO Advisor Podcast. We're doing something new. This is the CTO Advisor Podcast live. We're doing it via Periscope, via Zoom. So it's kind of some exception stuff going on here. On the other end of the line, I have the CTO of CloudGenera, Bobby Allen. Bobby, welcome back to the CTO Advisor Podcast. Thank you, Keith. Thanks for having me. Again, thanks for being a part of this really cool experiment of doing some live content around video and audio.

We're gonna continue to push the, I think the envelope here at the CTO Advisor. So we're gonna talk about a conversation, Bobby, that we've been having online with a lot of people lately, which is the repatriation of cloud workloads to the private data center. This has been a pretty heated topic the past few weeks, wouldn't you say? I would agree. It has been, and I've got some opinions about that, and actually some more customer data points, if you can believe that.

Things that some stuff actually came in hot off the press today. We love data on this show. So first off, define for us cloud repatriation. What workload repatriation? What does that even mean? So workload repatriation, in my mind, Keith, means that you had workloads that were running in the data center. You took them out to a public cloud provider, AWS, Azure, et cetera. And typically because you were scared because of the bill or because of the type of spin you had, you decide that you need to move it back to something that's more private and more predictable in terms of the spin.

So essentially public going back to private, typically things that originally were private is how I define repatriation. So from a size of happening of repatriation, one of our good friends in the industry, Corey Quinn, is obviously vested in making sure that enterprises move to the public cloud and they control costs associated with that move to the public cloud. And he's been a vocal kind of realist when it comes to repatriation. He's asked for like real data. What's really happening versus, you know, kind of this, you know, Keith, you're given kind of these isolated experiences of when it comes to repatriation.

Share with us some of your data. What are you seeing when it comes to repatriation? So I'm seeing a couple of things. So number one, there are people who are talking about repatriation, but there are also people that are actually doing it. And I think to talk about the first one, some of why people are thinking about repatriation is because they dipped a toe in the water, maybe doing some dev and test workloads, and running a few Linux things. And then they graduated to what I call the Cadillac workloads, running SAP HANA, running big, heavy databases, running things that were very data intensive, like Hadoop clusters.

And then what ends up happening, Keith, is between the premium storage and the egress, honestly, the juice isn't worth the squeeze anymore. Because it's one thing when you're consuming very economical services, like a small workload, like I said, a Linux workload, that are often lost leaders for the providers. A lot of people don't realize that. And then I'm going to go to run something like SAP HANA with terabytes of memory. That's a Cadillac workload that's offered at a premium. We have to pay attention, Keith, to where the providers are making their money.

They're not making their money on little one-by-two and two-by-four workloads. They're making their money off of those huge, massive terabytes of workload VMs. And when you're running those types of things, often people have not done the math. Because the cheaper stuff, the lower-end things are cheaper. They assume the higher-end things are also cheaper. And that's usually not the case. So depending on what you're doing, you're going to get stunned because you assume that the cost savings apply across the board, and they usually don't.

So I've heard that. But again, you say you have some numbers. Like, me and you have gotten into two conversations about this SAP thing, specifically SAP HANA, where companies have gone to HANA in public cloud, and they've gotten this shock in pricing. Like, you see the advertised price, and then you see the reality of what you're spending. You guys collect a lot of data. What areas, specifically, when it comes to, let's say, HANA, are customers realizing, wow, this isn't exactly what I thought it would be?

So when it comes to HANA, I think, again, part of what they're mixing, Keith, is the behavior of workloads that are more bursty, right? If you're running kind of what I call regular code on a VM, that's one thing. If you're running something like SAP HANA, in-memory databases that are large or almost never turned off, so you're running those 24-7. If you really plug that into a model like what we do to kind of project stuff out, you get burned up on those costs.

And if you talk about running that comfortably to running equivalent kit in the data center, those costs are nowhere close, is what we've seen, right? Because I'm going to go back to something that we talked about with Corey. Part of what Corey talks about is human behavior. And the reality is, I'm going to wade into the controversy a little bit here, Keith. When people talk about public versus private, is public always cheaper? There are some of our friends and some fellow influencers who have different thoughts about that, right?

I would say that a well-run, efficiently, well-utilized private data center is typically going to be cheaper than a public data center. But there are more options and more services and more variations on alternatives than a public data center. And where that breaks down, again, if you're running very, very large, very premium services that are running 24-7, that math almost always turns out that running that equivalent in a private data center would have been cheaper. Now, one of the things that I said earlier this week, Keith, that some people agree with and some didn't, I don't think people always go into public cloud either because of cost savings.

I think a lot of times they're going because of cost transparency. And what I mentioned was, a lot of these line of business executives, they wanna use a food analogy, you know, I do that a lot. They wanna pay for their own groceries. They don't wanna buy food for the whole street anymore. And so if I'm running something in the data center, but I'm covering everybody else's bill, they wanna take their chances, take their workloads or applications, go to public cloud and have visibility in what they're paying for.

But some of those same executives, Keith, like if you were a line of business executive, you're paying for everybody's dinner, so to speak. You're covering the tab for all of it. You take your SFP HANA to the cloud. Now you can see what you're spending, but you took a use case that is premium, right? As opposed to something where you could kind of prove the model out. So I kind of threw a lot of things in there, but that's what we're seeing.

People are making assumptions. They're jumping to the premium case and they're getting hurt as opposed to doing the math upfront, doing the calculations before the meter's running and realizing that those bills are really all burning them up. And they're material differences between public and private, depending on your behavior. So you highlighted a few different challenges that I've seen in enterprise. One is a behavior thing. Another one is when you try and move to a chargeback model or a showback model, you reveal some really interesting politics or finances within the enterprise.

One, you find out there's really big workloads or really important pieces that subsidize other parts of your data center. e. SAP HANA might be 30% of your overall IT infrastructure business or revenue. Whereas HR might be taking up 20% of the resources, but HR doesn't generate income. So to say that I'm gonna charge back HR for data center services is unreasonable. And this is where I think a lot of people who get into this conversation of public cloud versus private data center being more expensive is really missing the point.

This is a issue of perception. As a IT infrastructure leader, if I have a budget of a million dollars a year and you tell me you're AWS and you come in and you tell me, hey, Keith, your time to development will be cut down. Your facilities costs will be cut down. These things that's not in your budget today will be reduced. And you're gonna spend, your department is going to spend 10% more. 1 million on IT infrastructure. My organization might save $300,000.

My overall company may save $300,000, but my budget has gone up. And from a realistic perspective, I have to deal with that. My CIO isn't giving me a break. He isn't going to the app team and saying, you know what, Keith Townsend has saved you $100,000. Therefore I'm taking $100,000 from your budget and giving it to Keith. Organizations. I'm sorry, let me jump in there, Keith. I actually just had a very similar conversation to this with the finance team of a very large insurance company recently.

And so here's what happened. So two terms I want to unpack for your readers. There's unit cost and there's unit price. Those are two different things. Let me explain that. If I've got 24 servers, right? Am I charging for all 24? Am I charging for the 12 that are customer facing that are generating money, right? Because the HR servers, the shared servers, the things that are common things. So I've got 24 servers on the floor, but the 12 application teams that are customer facing are the ones that are paying really all the charges, right?

So that's price versus cost. So here's what happens. The 12 application teams that are paying for the 24, they've got the overhead of management, they've got the service now and all that kind of stuff. They're going to public cloud and essentially they're dropping the overhead of all those other things, right? Because they're just paying for the boxes on the software, not the other people, the bodies in the organization. So to get over there and what ends up happening, and this is part of why repatriation often happens, Keith, because to your point, when you get to the higher level, to the global CIO, not the line of business CIO, he says, wait a minute, I'm still paying for all the same people.

I'm still paying for the 12 servers that I can't charge back for anymore. But the 12 customer facing servers went to AWS. Now my bill has doubled in terms of what I'm really, really paying for, my green dollars, so to speak, that are going out of the door. And so sometimes, Keith, the issue is not that the math is wrong, the model is wrong. If you look at what each server actually costs, if you look at the cost of those servers, again, unit cost of the 24 servers versus the unit price, I'm paying for the 24 on the backs of the 12.

If I'm one of the 12 that's paying the bill, like if we went out to dinner, Keith, 24 people go to dinner, but 12 people split the bill. The 12 people that paid for it feel like they got shafted. And so if you're one of the 12 paying the bill, you want to pay for your own dinner because you're tired of carrying the other people in the business. That's really what's happening in some of these dynamics. So they say, look, I want to pay for my dinner, not other people's dinner.

I want to go to my own restaurant, pay for my own food. But then they realize they're not just, they actually went to like a Ruth Chris and they're paying for a premium steak now. And so now they want to go back to mom's house and sit at her table and eat her food because they realize it's a different world out there. So sorry, I'm hoping about making people hungry on the podcast. Hope you had lunch already. But this is part of what happens when people talk about the models.

There's a difference between the math and the model. The cost of a server is different than the price of a server. How are they recovering the cost in the enterprise? And if you miss that, we're going to miss what's driving a lot of the behavior. So let's talk about that reactionary activity once you figure out that, hey, you know what? I was in mom's house and mom was covering part of the cost. So there's the price versus the cost.

The price of my rent to mom was $500. The actual cost is much more than that because mom was subsidizing my living. So honestly, for some people going to public cloud, for some of these workloads, and you mentioned some of the lower end workloads, this pays off. Like the math is really simple. The storage is much cheaper. Compute can be cheaper, especially for simple workloads, especially if I'm using the elastic nature of public cloud and I'm spinning up, spinning down, et cetera.

I'm realizing a lower cost and a lower price for some. But then there's some that's realizing a higher price to them. Is it that we're seeing entire organizations repatriate workloads? Are we just seeing the reaction to some of that higher price activity come into play? So what I've seen, Keith, is that a lot of business executives typically want to go out to public cloud because again, they're tired of covering the overhead of other people's projects and other people's dinner, so to speak.

But when it gets to the top of the house, the top of the house wants to repatriate because they're not saving any money. They're paying for all the same stuff in the data center. They're now paying for additional stuff in the public cloud. And so now they've got two different sets of costs. So let's use a different analogy. If I've got a college student, Keith, who just went away to college, I don't have a lower mortgage because my son or my daughter moved out.

I'm paying for my place and I'm now paying for their place as well. They've gone off to college and so I'm paying for maybe a dormitory in my house. I've still got a data center that may have empty space but I'm paying the same bill. When they come back home, and that's what ends up happening with a lot of parents, Keith, they say, if I'm gonna pay X amount to the university of such and such, you're gonna come back home, you're gonna repatriate because you're gonna move back into your room because I'm already paying a bill for the college tuition.

The same thing happens. The individual who wants the choice wants to go where they've got access to the most options and the most choices. But the person actually paying the bill, the global CIO or even the board, they're gonna say sometimes, especially if you're not earning, right? That college student is not making any money, Keith. They're not contributing to the Allen or the Townsend household. They're dragging bills out of that. But if they come back home, then I have literally reduced bills.

So repatriation, the first part of repatriation, people going out the public cloud, I think is driven by choice. But I think people coming back home is driven at the top of the house because they're tired of the fact that they're not realizing the cost savings that they were promised. So I talked about this a lot with HCI when it first came on the market, which was there's this idea that if I moved to HCI, I'll save on storage admins, I'll save on compute admins because those OpEx charges collapse.

In reality, what happens similar to your antidote that you just gave is that costs just increase. Now, instead of having my Dell EMC VMAX array in my compute systems, I now have my Dell EMC VMAX array in my HCI platform. So I don't get rid of my storage admins. I get storage admins plus my HCI admins, and I don't save costs. I may be more agile, but there's no doubt about that. I gain agility. Like the speed to value story is absolutely there, but the cost savings isn't necessarily there.

So you're saying that unless you're all in on public cloud, unless you've left the data center, unless there's a value prop for the HR applications, the finance and accounting, all these back office systems, the exchange service, et cetera, et cetera, the stuff that can go into SAS or whatever, unless you have a full strategy for exiting your data center, you don't necessarily realize those cost savings. I agree 100%, Keith, and that's a conversation we have a lot too. The data center is like a commercial property, right?

If you and I were landlords and we had 10 tenants in a property, if nine tenants move out, Keith, I can't get rid of that property yet. That last tenant in the data center is forcing me to hold onto that property. And what ends up happening is here's the way the math often works. You and I are landlords of that data center. We're charging our 10 tenants for that money. Every time a tenant moves out, the price goes up for the people that are left.

You do not wanna be the last person to turn the lights out because now you're the last tenant paying for that entire data center by yourself. So that behavior, the model, Keith, not necessarily the math, is pushing people out of the data center. But if you're the person paying for the building and then now paying for where those tenants moved out to, your bills have gone up. So I agree 100%. You have to have a holistic, often multi-year strategy. When are we actually gonna be able to turn things off or sell that building or abolish the property?

Because until you can do that, you're gonna be paying more money, not less. Yeah, I've had this conversation time and time again. I've been in this situation. I've gotten out of a data center. We've done data center consolidation. Just go back to facilities and say, hey, facilities, here's the data center that we had in Anaheim. We no longer use it. The facilities VP says, you know what, Keith? Yeah, that's great. You're still gonna pay me for it, but yeah, it's great that it's empty.

I don't have another customer inside of the enterprise to take this space. I'm not going to go and tear up the raised floor that you had me build, all this electricity specifically for a data center. That's your cost for the next five, 10 years until I can get someone else, one, to pay for the renovation, the capital cost to pay for the renovation. Then two, the OPEX cost to take that burden away from me because our CFO isn't giving me more money just because IT left the data center.

Exactly. I think you hit it, Keith. If there's not elasticity on the cost recovery side, that's really the issue. When do I actually get some money back? When do I get a rebate, so to speak, or value kind of coming back to me? Because in the meantime, and that's actually the issue if I could be opinionated for a minute, Keith. That's the challenge that I have with some of the vendors and the CSTs. Their models are overly simplistic.

They're assuming that just because you move one child out, that now my mortgage is going to go down. In the real world, that does not happen. I can't cut off 10% of the data center or 25% of the floor space. It's really kind of all or nothing. If I'm not all in, or eventually getting to the point where I can turn something off, and so we need to be more honest as we're advising these companies, Keith, then in the meantime, you're going to be paying for multiple residences until you consolidate.

So in that one, two, three year interim period, your costs are actually going to be higher, and then eventually they'll get lower, but by that time, you might be in a different job, or the company is in a different direction. So to recap the conversation, no, companies are not wholesale leaving public cloud. What they are realizing is what I've seen is that when they move SAP HANA to the public cloud, and they're paying for one of these 24 terabyte RAM systems that run 365 days of the year, plus their inability to right size the workload to begin with, because I'm not going to, this procurement just doesn't work like that.

IT does not work like that. I can't procure a 16 terabyte solution, and then go back to finance and say, I need a 24 terabyte solution, then come back to finance and say, now I'm back to a 16 terabyte solution. One, SAP HANA is not elastic like that. So you're going to provision what you think you'll need for the next year or two years based on how I would do in the data center, because that's just the way these applications are designed.

So companies are realizing that, and when they realize that, they're not wholesale moving out of the public cloud. That's not what we're saying. We're saying that workloads, that there's obvious value from a speed to agility perspective, that I can, a speed to value perspective, I can get value in business advantages, competitive advantages by moving these applications to a public cloud, or building these applications in a public cloud. Those things stay. But for these things that kind of keep the lights on, SAP, SAP HANA, Oracle databases, et cetera, where the math just doesn't make any sense, if I'm not moving out of my data center completely, those things, I'm seeing those things come back to the private data center.

And Bobby, you and your company are seeing that math in reality. Let's do a pitch for CloudGenerate as we wrap up. What do you guys do, and how are you guys getting this type of data? Thank you, Keith. So CloudGenerate is a company based in Charlotte, North Carolina. I got to represent for Silicon South, as we call it in the Southeast. And we're a company that helps with workload placement and transformation. So when you want to model the art of the possible and hypothetically map it to public or private venues, we can simulate all of that stuff for you before the meter's running.

Containerization, serverless, public versus private stuff. If you want to model it before you're paying for it, that's what we do. Showing the art of the possible quickly. com. You want to learn more about me, you can find me on Twitter at CTO Advisor on the Twitters. Until then, you can catch us in iTunes. Make sure to rate us. Talk to you next CTO Advisor podcast.