Podcast: unpacking the evolving challenges and decisions surrounding public cloud strategy

Sharing expert insight into public cloud strategy...

In the latest episode of Asanti's “In Conversation With” podcast, Asanti’s Emma Lauchlan speaks with Chris Talliss, Director of Cloud, Data and Apps at Wavenet, to unpack the evolving challenges and decisions surrounding public cloud strategy.

Drawing on Asanti’s recent research, which found that 52 percent of UK organisations plan to reduce their reliance on US-based cloud providers, Chris discusses the two major drivers of this trend: data sovereignty and data residency. In this insightful episode, Chris shares:

• Why hybrid and private cloud strategies are growing in importance
• How the Cloud Act and global politics are prompting a rethink of cloud-first strategies
• The danger of falling into “modern legacy” through well-intentioned cloud migrations
• How AI is reshaping infrastructure needs and driving demand for colocation

Why understanding your workloads, not just your infrastructure, is critical. Whether you’re re-evaluating your current architecture or exploring how to scale AI workloads securely and cost-effectively, this episode offers valuable guidance and perspective. This is a must-listen for any IT leader navigating cloud transformation.

Read the latest research here: https://asanti.com/press-release/uk-it-leaders-shift-cloud-strategy-over-data-sovereignty-concerns-new-research-finds/ 

TRANSCRIPT

Introduction: This transcript has been generated for reference and accessibility, with subtitles included on the video for easy navigation. It will not be 100% accurate but should be very close to the conversation.

Hi, welcome to the next edition of In Conversation With. Today I’m joined by Chris Tallis from Wavenet, who is their director of Cloud, Data and Apps. Chris, welcome.

Hello.

Today we’re going to talk about public cloud strategy. What, how did we get here? What are the challenges? And we’ll talk about some recent research that Asanti has conducted. The key headlines start from the research – 52% of UK organisations plan to reduce reliance on US-based cloud providers. Two key reasons, data sovereignty and residency. What are your thoughts on that?

Very interesting. Okay, so let’s start with cloud, data and apps. If you’d asked me that question a year ago, I would have said cloud. Probably the first thing to understand is why we’ve moved to cloud, to cloud, to data and cloud, data and apps. Customer value.

Our role as an MSP is to help customers release value from the technologies that they invest in. If you now extend that into public cloud, cloud first strategies, cloud native deployments, cloud native applications, using third party clouds to centralise data, harmonise it, analyse it, use AI to interrogate it, maybe monetise it, et cetera. That whole part of, in your description, public cloud strategy has obviously been on our lips for years, for many years. And I think, pertinent to probably the last 12 months, and certainly the arrival of said act, is the blend between public and private. So, absolutely from our customers, we find exactly the same thing as what you said in the analysis.

Absolutely this hybrid approach. And sometimes that means adopting public and private. Sometimes that means adopting more of one and less of the other. But nonetheless, we’re absolutely finding a blend is the way that organisations are going, but not through technical prowess, more through pursuit of value.

Yeah. So, you mentioned pursuit of value. I think our previous research showed that 91% of businesses were repatriating out of public cloud due to cost. How much do you think Donald Trump 2.0 is now playing on the minds of that 52% who say, actually it’s not even about cost, it’s just about sovereignty and residency.

I think it’s amplified it. Yeah. It’s absolutely amplified it. And ultimately, as with the march of technology, it’s very rarely one thing that causes those events or those changes in trends. But what I think we’re seeing this time is technology specifically being aligned to economic trends. And obviously we know that economic trends take the money markets with them and therefore acute interest when technology follows that same way as well, as we are all consumers of said technologies, you can’t ignore that. So, I don’t think there’s an IT leader out there that doesn’t have this or a flavour of this in mind. And I think the hyperscalers have done a lot to help with that thinking. You’ll recall that less than a year ago, the narrative was first step centralise all your data, give it to a third party cloud provider, that’s your job one, was always sceptical, private cloud providers and the co-location providers, they were very much more around, you don’t necessarily have to move the data, you just have to be able to access to the right point in time. To my point, it turns out a hybrid approach might be appropriate.

Yeah. And all of those businesses that are moving or considering their technical setup because of the US Cloud Act, we’re seeing providers like Amazon reacting to this by creating a European cloud. I mean, the fact is that that’s still an American technology no matter how many times they put European in front of a building name or a job title.

Yeah, we’ll take the example of AWS with the GovCloud, that’s a sentiment there eating one’s own dog food, that doesn’t take GovCloud out of the umbrella arch of the act. The same is true.

For European data centres, et cetera, the fact they are owned by US organisations is sufficient.

No further amount of segregation or otherwise helps with that. But one thing I think we can see from history, which I always find acutely interesting, is rather like financial markets, technology markets also automatically correct.

So as for example, threats to data, in this case, we’re talking about data sovereignty and the US government’s rights to access that data irrespective of jurisdiction. So, what will that trigger?

Well, march of private cloud, yes, of course. And maybe the march of technologies to one time encrypt data, key technology for things like HSMs, which prevent the hyperscalers from being able to decrypt the data even in the event that they have to surrender it. So, one of the things I love about technology, course correct.

Yeah, certainly not an industry or a sector that stays still, right?

Correct, exactly. So, you mentioned earlier about cloud first being a strategy. How did we get there? Why did the UK government say, you know, we’re going to be cloud first, this is it, this is the way to go? What led us to that point?

A series of events. My observation was we had been in static infrastructure for quite a long time. So, the arrival of single click architectures, deployments and being able to describe infrastructure as code was incredibly appealing. So that obviously arrived at a point in time when infrastructure had reached a point where it couldn’t legitimately be automated much further. So that was fascinating. Second thing I think was the concept of cloud economics. And certainly, from our business, cloud economics is one of the great things that’s come out of third-party cloud. We use that for our solutions and services models as well for professional and managed service. We use the concept of cloud economics. So that came and increased the drag.

And the third thing was you got to bear in mind at the point at which cloud adoption and cloud first strategies were really booming, the hyperscalers were spending a fortune incentivising organisations to adopt their technologies. And whilst that created a concept of, you know, essentially building out, getting started and then expanding from there; by funding those initiatives, it created a ton of bad habits and created a concept which I heard quite recently of modern legacy.

So, you think cloud as being a new concept, a new thing, but guess what? The way it was adopted by well-intended, I hasten to say, but by well-intended funded initiatives immediately created a modern legacy.

Yeah, so I think, you know, from a customer perspective, they’re looking at this new at the time public cloud, it’s quick to spin up like you say, and there’s no upfront capex. So, it looks on paper first pass like a win-win. But then as you start to, yeah, as you start to spin up, as you start to transfer data from the point at which you’re kind of generating it into the cloud, then back, that’s when the costs come, right?

Exactly. And that is when the lift and shift fallacy bites. It’s ever so easy to lift and shift. There’s tons of software providers out there that’ll help you do it at the click of a button. All you end up doing is taking the same workloads, obviously, into a third-party cloud. All you get is something that’s a bit more expensive and a little less good than where it was before. And lo and behold, exactly the finding at the top of your paper finds exactly the same thing. So it isn’t purely to do with the tethering to the act. I think it’s a multi-dimensional puzzle, this one, based on customers moving to a thirst for value, as things like FinOps become more important, an aging of cloud-first strategies, and then an application of one legacy.

Yeah, so we could be getting to that tipping point for a number of years now. People have realised that that public cloud adoption wasn’t going as smoothly for whatever reason. It means maybe perhaps that they couldn’t take the old legacy apps and put them into the cloud, so they’ve been battling that for years. Then the CFO gets a hold of their bill and goes, wait a minute, this is way higher than we were expecting. So, with that sort of data transfer or data taxi that it’s sometimes referred to, why did people not foresee that? Is it because they didn’t know how much data that app was going to use until they went into that cloud environment? Why was that not? And how is it still not an identified cost centre?

Most of the organisations I talked to say it was purely done through convenience. A symptom of business pressure. And bear in mind, I can say this, over a relatively short number of years here, the role of technology has been to move from a cost centre to a profit centre. You look at the reason that organisations are investing and chasing the AI/data path is to monetise it, is to monetise it, that’s why. And therefore, taking that, taking the path of least resistance, suddenly centralising your data into a hyperscaler looks like the perfect option. Why wouldn’t you?

Except then you tethered to it. Two things happen. Number one, something like the Cloud Act shows up and gives you some indigestion. But then problem number two shows up. What happens if you can’t operate or experiment on that data as much as you’d like?

Operating as service platforms for interrogation of data is brilliant as long as you are on the same gauge of railway as the hyperscaler is on. If you’re not, you’re struggling.

And you mentioned just a moment ago, private cloud. So, explain what the difference is between that public cloud environment that we’ve been talking about that’s kind of in the arena of the Trump 2.0 and maybe the government can reach in there the public platforms. And what is a private cloud? And what are the key differences?

So, some of the differences are quite subtle actually. And what I’ll do to answer that, I’ll describe the Wavenet private cloud portfolio, which is essentially a multi-tenant private cloud, which is our own platform, which is multi-tenant by nature private, i.e. private between us and the end user customer.

Very difficult to segregate that from third party cloud. The only difference being obviously different service reps, et cetera. Second point is dedicated. So dedicated infrastructure, be that customer-owned infrastructure or a more complex mechanism, such as GreenLake, for example, an excellent way to get access to private cloud technologies.

And then the third one is colocation services. Now, the colocation service, I think, is a particularly interesting one at the moment due to the march of GPU adoption, where different shaped racks with different cooling and power requirements in order to scale them suddenly become of interest. And whilst for a lot of the organisations, for example, that Wavenet work with building their own AI models or becoming a model builder is not really in their swim lane, consuming them is, and therefore getting access to services and facilities that allow them to do that in a cost-effective way creates a space for private cloud.

Yeah, so private cloud is essentially, when we talk about cloud, right, it isn’t a floaty thing in the sky. It really is, whether it’s public, whether it’s private, it’s just a data centre that we’re talking about. Exactly, that has been the case for a very long time. And as I say, when I describe it, I’m public, private, and colo. Those, to me, are my three compartmentalisation approaches.

And a private cloud would normally be run in a data centre for Wavenet customers in the UK? Yeah. Yeah, it would.

So, when we think about the adoption of cloud, we’ve talked about costs and things like that, but what’s your advice to organisations if they are starting to see those charges creep up, if they’re slightly starting to worry about the cloud act? What’s your advice to those organisations and how can Wavenet help them?

My advice, quite specifically, take the time to understand your workloads. There’s a concept called application mobility, which was around a few years ago, and the concept said there’s never a great point in time to invest in understanding your workload, maybe automating part of its orchestration, automating its deployment, automating its infrastructure, et cetera. But that said, when you do so, it gives one the ability to operate more effectively with that application. As new things emerge, as new trends emerge, et cetera, the understanding of the workload, as opposed to the infrastructure the workload sits on, it’s a subtle difference, is absolutely fundamental to being successful. And that is true not just from the strategy perspective with an organisation from a snapshot, but also from the pillars that make up the delivery of that strategy as well, right? So there has to be a point when you understand the workload as opposed to the infrastructure it sits on. And that was at the crux of the third-party cloud adoption that took place, particularly if you look at the technologies that did virtualisation, lifts and shifts, those took that same construct with its frailty into a slightly more expensive realm. If those workloads weren’t understood then, cost are likely to have gone up, maybe now is the time to have a look at them. And modernising them doesn’t necessarily mean tearing them all apart and rewriting them. There can be varying degrees of that. That is a scalar event.

And again, an area specifically where Wavenet has the expertise and the size and scale and the impact capability and capacity to go across the cloud and cloud infrastructure, data and the applications that drive the production and analysis of said data.

And in your role with Wavenet, are you expiring saying that customers are coming to you more for this help we need to re-architect here? Or are you getting the opportunity to set them on the right path from the offset? Or is it a mixture?

Huge variety, purely because of the types of customers we get the luxury of working with. So, because we have our products and services across the board, as you’ve probably seen from our portfolios quite wide, what that means is that we resonate with a lot of different organisations across tons of different verticals. And that gives us the most fascinating problems. But to answer your question more specifically, what we do see is customers falling into one of three modes.

And that being those that are almost exclusively legacy. So those organisations where they are, for whatever reason, are in states where they’re running legacy applications and legacy infrastructure. And for them modernising that as a material endeavour and one that might be outside of reach for a variety of reasons.

And that legacy infrastructure might be on-prem?

Yep, mixture of the lot. But often quite physical in nature, including things like the old classic in-building data centres, those sorts of things. Then you have the next mode, which is those who have done a degree of modernisation to that infrastructure. This is where the majority of the organisations we work with are. And our solutions and services that wrap around that aren’t just around change and run; they’re also around the advisory services as well. Because they’re the ones who could gain the most but are also the ones who are struggling with the modern legacy, they’ve created by moving from the previous mode, if you see what I mean. Then the final one is those that are innovating directly into third-party cloud. Now, obviously, they’re the ones where something like the Cloud Act is particularly important, do they continue to do that?

You know, is third-party cloud the right location for them to be modernising in? Probably yes, but I think things like the cloud act do force our hands to be critical of that and make sure that we’re not doing that purely through history. So those are really our three buckets. And we help in different ways across advisory and professional services and on-demand as well.

Yeah, and do you find across those three buckets that there is the same size of organisations or the same age of organisations? I think with public cloud, we mentioned that there’s no upfront capex, so that would lend itself well to a start-up perhaps or someone that may be just spinning up into environment for adding deep purposes.

Yeah, absolutely. So, the majority of the innovators and the ones who can, in our collection of customers, tend to be those smaller, more agile organisations that enjoy that benefit. But that said, I think the subset of that that’s really important is those parts of a larger organisation which operate in that way. So, when you use the expression start-up, I think it’s important that everyone understands, you know, just mean anatomic organisation as a start-up. It can also mean another part of a business where they’re bootstrapping a new idea or innovating something or taking something hastily to market. All of that falls into that startup category as well.

So, thinking again about, you know, these organisations that may have gone into the third-party cloud as early adopters or because they were a relatively new organisation, if they want to come back out of these services, out of these AWS, Azure, where are they going to go? Right.

And in terms of co-location, so is that an option that you’re seeing customers take more, more and more?

Yeah, absolutely. Curiously, there’s two things that I’m seeing driving our customers’ adoption of co-location. Number one is the pressure that’s being placed on facilities, space, the adhering to standards and such. That’s a big problem for a lot of organisations. And therefore, keeping data, even in a private context, outside of co-location facilities is really hard to do. Second thing is the curious makeup of some of the workloads that we end up rehoming.

These can be a whole variety of things. But again, GPU is a great example. The standby power consumption on collection of Nvidia cards, it’s relatively high. The size and form factor in terms of how to handle them, liquid cooling, et cetera, et cetera, makes perfect use case for co-location where the size, shape, form, function of the space can be completely customised by the customer.

I think the final point on that is the concept of cloud economics. So, if you recall earlier on, we talked about the concept of cloud economics bringing into the market a new consumption-based model for IT and for technology generally. I think co-location services in adopting a similar model for power, for example, has been an absolute transformation. And that has been stimulated whether we like it or not by the hyperscalers introducing the concepts that cloud economics taught us, how to manage it, budget it, consume it and not be wasteful with it. And that concept has pulled straight through into colo, which is a fantastic improvement in the market. Yeah, okay.

And then finally, let’s chat about AI. What is your opinion on how this is going to influence infrastructure decisions? I mean, what does it mean to Wavenet and its customer base? Are you seeing influence decisions or?

Yes, 100%. It comes up in every conversation, comes up in every meeting. But it also comes up in our experience with an interesting twist as well, which is ironically, birth from one of activation. A lot of the conversations that we’re having with our customers are based on the need, the proof of value, proof of concept philosophy of a traditional old school business case. I would like to buy some infrastructure to do some AI-ing on this data. Brilliant. How much is it going to cost? X amount, okay. Can you find that in the first instance also?

And then how are we going to recover that cost? We don’t know. Why don’t we know? Because we don’t know what we’re going to find. Or can’t we use prior art? No, not really. The technology is still relatively too new. We’d be guessing.

Okay, how do you unlock that? And that conundrum is very current for the majority of the organisations we work with. Truth is, for quite a while now, we’ve been working with customers along the proof of concept, proof of value, low-cost type models, not to enable them to step straight in to AI but enable them to find the golden threads and the value chains in the data.

A simple exercise of data discovery to find what you’ve got paired with an on-demand model, in our case, to explore the art of the possible, to create an infrastructure environment, even if it’s relatively small, and then to see what value is there and what value can be determined from the data.

And I think this is where the old classic myth-busting steps in, surely you need an army of GPUs to be able to do that.

Yes, of course, if you want to do it for millions of users. If you want to do it for three users, you can do it on a CPU. Fine. Okay, well, surely the software stacks are incredibly complicated, and I’ll need an army of data scientists to do it. No, not at all. That might have been the case a year ago, but again, as I said before, I love the fact that technology market’s course corrects. So surely all of these tools that were very difficult to deploy and manage and dependencies, et cetera, were really tricky a year ago. Yeah, no, they’re not now. They’re just not.

Are your customers, looking to develop AI, or is it just looking to deploy AI-enabled applications?

Right, two different types. So, there’s a big difference. Massive difference. I see it’s two different types. You’ve got your general-purpose AI, which is your co-pilots… Your deployed applications. Yeah, yeah, yeah. Your chat GPTs, et cetera. Your deployed applications. Then there’s the integrated AI.

And I think when you start to look at the target, the target for most of the organisations is monetisation. That takes you off down the integrated route straight away. Now within that, I’m deliberately glossing over the distinction between generative and predictive and all of that. Just ignoring that complication for a minute.

There’s still a ton of organisations that we talk to that are struggling to start. And start means get going to prove there’s a business case. And that shouldn’t be that hard. And that, again, is where Cloud Economics has helped us. So, it’s taught us a model that enables us to take our solutions and services into that puzzle.

We call it on-demand data, to enable those customers to start and to find out. To start developing AI. In a private cloud. So, they don’t have to worry about centralising their data first. They don’t have to go through an endless stream of data repatriation. It can be done behind closed walls, get started to see how it goes. So, is AI featuring a lot of conversations? Yes, but mostly in the degrees of struggling for business case territory. Not because it’s not there to be had, but it’s the degree of certainty on what you’re going to find. So, you get a Mexican standoff.

When we discuss the two types of AI, and Asanti, I’ve spoken to people that have developed AI applications, and they talk about the two sides being the learning. So, the large language models that are being fed these algorithms. That’s where we’re seeing the high consumption of power, the GPUs that you mentioned. And then like we just discussed, the deployed applications just become like any other application.

And from people that I’ve spoken to, they’re using probably third-party cloud to quickly spin up, do a bit of learning and then come back out of it for AI development.

Yeah, that’s an interesting one for two reasons. One, because if you actually do that as an engineer, one of the things that I’ve found in doing that with the third-party clouds is you only can access certain facilities, which means you are to some degree tethered to that particular hyperscalers interpretation of what it is you’re trying to do. If you hand crank your own one, none of that is the case.

So, it’s a bit more control. So, it all depends on your underpinning hypothesis and whether you even have one. The most successful projects I’ve worked on in that space have not had a hypothesis. So, they haven’t been working to an objective, and they’ve been done in a private context, which has enabled them to explore and to play. That is incredibly difficult to do behind an abstract. Whilst it’s incredibly easy to consume behind the abstract, that can often mask your own findings.

Okay, so just wrapping up then, if you could give one piece of strategic IT advice to any leader that’s planning their roadmap for how they’re going to host their critical infrastructure, what would that be? What would it be?

At some point, you’ve got to properly understand your business, understand your business services and understand your workload.

All the time you choose not to do that, it becomes incredibly difficult to set out your strategy properly and you end up getting blown from public to private, from private to colo, from colo to public, round and round it goes. Ultimately, you have to create the space at a point in time to understand what the thing does, what value it is to the business and how it works in the context of the business. Then you can start to take some decisions about where it’s best placed. And also, I think the other thing on that that is critical once you do understand it, revisit it regularly.

Tech moves on super-fast. That’s why we’re all in the industry. Don’t try and pretend the decision you took 12 months ago is now still… Roll with it. Exactly. Yeah. Roll with it. Yeah. Revisit it.

Brilliant. Thanks very much, Chris.

Thank you very much.