WTF Is Cloud Native, strategy

WTFinar (with transcript): When (If Ever) is Cloud Repatriation the Right Move?

With a few notable exceptions, the traffic of workload movement over the past decade has moved firmly one way: from the data centre to the cloud.

But in 2016, Dropbox surprised the industry by announcing that it was going against this trend by investing over $50m on non-hyperscale data centres.

And so the recent public announcement by 37Signals has rigorously reignited the debate in this area; reiterating Andreessen Horowitz’s frequently referenced quote: “You’re crazy if you don’t start in the cloud; you’re crazy if you stay on it.”

Is this the wave of the future, is it already happening, and why will it take off, if at all?

In this event, we try to resolve the paradox that some experts say companies are ‘crazy not to repatriate’ while others say repatriation is the ‘exception, not the rule’.




Ian Miell, has been an engineer for over two decades, and was an early Docker user and advocate, co-writing Docker in Practice for Manning, and giving various talks on it. After spending 14 years writing and maintaining sports betting back-end systems at OpenBet, he was lured away by Barclays to help run their OpenShift platform for a few years before joining an abortive cloud native transformation project at State Street. He has written books on Git, Bash, and Terraform, and given talks on these and related subjects at various conferences.

Jamie Dobson is co-founder and CEO of Container Solutions, a professional services company that specialises in Cloud Native transformation. With clients like Shell, Adidas, and other large enterprises, CS helps organisations navigate not only technology solutions but also adapt their internal culture and set business strategy. A veteran software engineer, he specialises in leadership and organisational strategy, and is a frequent presenter at conferences. He is also the co-writer of the O’Reilly book Cloud Native Transformation: Practical Patterns for Innovation (currently available for FREE)


(Skip the intro and jump straight to Ian’s talk).


Jamie Dobson: Welcome, anybody who's new to these chaotic events. My name is Jamie Dobson. I am the Chief Executive of Container Solutions. But one of the nicest parts of my job is I get to do things with the community, either in person at conferences. I occasionally write. I did write a book once. If you need to go to sleep, if you've got any troubles with insomnia, buy a copy of that book, your lights will be out in two minutes. And I get to do these WTFinars with my wonderful friends and guests. Today's WTFinar is with Ian. And we're going to be talking about repatriation. Now, to those people who are not new, you know what's about to come next. There is going to be a small mention about housekeeping. We're going to make some announcements and then, of course, we're going to move into the main event.

We do usually have a gossip section. There is no gossip today, but I made up for that with my joke about Ian and magicians. I hope that went down well. Housekeeping, we mainly focus on the code of conduct in this part. Carla will share a copy of the code of conduct. It tells you all to be reasonable, respectful, and kind to each other. Please do take a look at that, if you haven't seen it before. We do enforce the code of conduct. Breaches are taken seriously. You can bring any potential breaches to my attention, to Carla's attention, and it will be dealt with. And if, in an extreme case, we have to, people will be removed from this event. We're doing this because the space, the technology space, it wasn't always safe for everybody, but we are part of a bunch of people who are trying to make it safe and inclusive for everybody who is interested in technology.

So this is why we have a code of conduct and this is why we take this stuff seriously. Announcements, yes, you can subscribe to our newsletter. We're the least spammy newsletter people on the planet, in the earth, oh, I can't speak anymore. We are non-spammy. If you scan that QR code, you can get access to upcoming events, free content, an amazing newsletter. It's the envy of many people in the industry. And if you're interested in a job, you can click this box that's says this, I'm interested in a job. If you're interested in buying something, you can click that box as well, but you don't have to. You can just subscribe at the level that you are interested in.

And if you don't click any of those other boxes, you won't hear anything about jobs, potential work, et cetera, et cetera. We're proud of our mission, which is to educate as many people about Cloud Native as possible. This does mean that we also do in-person events. We have a live panel in both Amsterdam and in London. Now I get the feeling I should know a little bit more about these events. Carla, what are these two events?

Carla Gaggini: Well, actually, these events, we're very excited about them because they're live conversations about DevEx, developer experience, and it's a topic that we saw getting more and more traction in the past months. We covered a WTF issue as well. So at the moment, the date in London is being rescheduled because there is a rail strike. Of course.

Jamie Dobson: I've just heard the rail strike has been canceled. I heard that five minutes ago.

Carla Gaggini: What?

Jamie Dobson: So you might need to reschedule your rescheduling.

Carla Gaggini: Oh, that's great. That's great.

Jamie Dobson:

Okay, so these are in-person live events. And who is hosting these events from our side?

Carla Gaggini: Well, from our side in London is one of our super engineers, Robel. Who's also going to be hosting one of the tracks of the conference. The WTF is SRE conference on DevEx. So it's all DevEx journey.

Jamie Dobson: All tied together, exactly. Robel will be hosting the one in London and then he'll be appearing at our conference. I'll mention that in a second. And the one in Amsterdam will be hosted by?

Carla Gaggini: Cari, another lead engineer of CS.

Jamie Dobson: Another lead engineer. We've been talking a lot at Container Solutions, recently, about engineering led marketing and engineering led sales. This is kind of what we mean by that. What that means is, at those events, you won't have to put up with anybody like me. You'll get to meet an actual real life engineer who actually programs things. I haven't done that for about 15 years, much to my regret, actually. All right, speaking of WTF, we do have a conference coming up. You can sign up now. It's going to be awesome. This is running from the 4th to the 5th of May. It's the first time, I think, the first time we've come back in person with the conference. Carla, is that right?

Carla Gaggini: Yes, that's correct.

Jamie Dobson: Since the pandemic. Amazing speakers. Check out the website. Brilliant speakers, loads of brilliant attendees, lots of brilliant sponsors. And I know what you're thinking if you're watching this, should I sponsor, shouldn't I sponsor? You should probably sponsor, right? You get brilliant bang for your buck because it's a nice conference and all the cool people come along to it. And that is my, no, it's not my final announcement. Oh my book, the book. Listen to this. A fantastic book about Cloud Native, DevOps and modern software. What an amazing review.

The book is free to download. Now, it's not free on Amazon, it costs you about 50 bucks. So if you want a physical copy, go ahead and get it. However, if you'd like a free copy of the book in PDF, go ahead, sign up or stick in your details on the website now, and you can get a free copy of the book. Many people who went cloud native, if going Cloud Native is even a way to describe it, have done so successfully with lots of the patterns, and guidance, and tips and tricks that appear in that book. So we're quite proud of that. It's not really my book, it was written by three people, and I did the smallest part. But I was very proud to be involved, and there's basically at least 10 years worth of best practice baked into that book.

And now the main event, David Baddiel, himself. Ian, David Baddiel lookalike extraordinaire.

Ian Miell: You'll pay for that. You'll pay for that, Jamie.

Jamie Dobson: I know, I suspect I will. I've outed you. I'm so sorry. How are you, mate?

Ian Miell: All right. Not bad. Yeah, I get people in the street shouting that at me, so I'm quite used to it. Thank you.

Jamie Dobson: I'm going to stop sharing now, Ian, so that you can share and then step the people through the questions. I think I've actually, have I stopped sharing already?

Ian Miell: Yeah, I should be sharing now.

Jamie Dobson: Perfect. No, you did that seamlessly. Professional.

Ian Miell: Until you mentioned it.

Jamie Dobson: In that case, it's over to you.

Main talk



Ian Miell: All right, Jamie, thank you very much. Welcome, everyone, to this seminar on when, if ever, is cloud repatriation the right move. It'll be great if you can put stuff in the chat during this. I really welcome that kind of engagement because this is particularly, it's a topic that gets people's blood boiling, one way or the other. So yeah, please do that. This came up recently. This topic came up recently again, and it surfaces every now and again. Though about two weeks ago, I published a piece on it on our blog, which was quite well received. And as a result, we thought we'd do a webinar on it to get some live response to it. So let me actually, I need to activate my chat. How do I do that?

Jamie Dobson: Now, Ian, is it fair to say one of the reasons we wanted to do this... First of all, we don't have a horse in this race. Do we? We don't sell repatriation services. It's not something we're pushing. So commercially, we're interested in everything. We move people to the cloud, we can move them back if they'd like. But this topic is divisive, isn't it? And that's one of the reasons we wanted to poke it a little bit.

Ian Miell: It's very divisive. Sorry, I'm still struggling to find how I can get chat up while I'm looking at it.

Jamie Dobson: Shall I send you a message, and then maybe it will appear?

Ian Miell: Yeah, maybe. Yeah. It is a very divisive topic, and I'll come onto this in the talk, but also, we don't have a horse in this race. Most people think, because we're a Cloud Native consultancy, that we're pushing cloud on our clients, and often that's not the case. Often, we tell them not to, that we don't think it's a good idea for them to use cloud. Cloud Native, of course, doesn't mean use cloud service providers, use the hyperscalers. It can also be on-prem, and we even have customers who are not on the internet in production. But yeah, it is a subject that comes up with our customers every now and again, but it's not something we're particularly invested in. I still can't get chat up, I don't know. Well if you're chatting to me, I can't see it.

Jamie Dobson: I'll monitor the chat, Ian, don't worry. I'll do it for you.

Ian Miell: Can you look out for me? Thank you. All right, let's go on. Okay, so the talk today, I'm going to be covering the following topics in roughly the following way. We're going to talk about why it's a hot topic, who's done cloud repatriation, talk a little bit about cloud refuseniks, people who are not going to the cloud. Talk a little bit about who should consider it and who maybe should not consider repatriation. And then finally, we're going to look at what's next for the whole topic of repatriation and cloud, in or out.

Jamie Dobson: And I would just like to say at this point, Ian has specifically asked me to be obnoxious and interject today to get the questions and conversation flowing. He really did ask me that. And I said that's easy, because that's what I'm like normally. So if anybody's got any questions, get them in the chat, and if you really feel compelled to scream out, we think you should go ahead and scream out. Is that right, Ian?

Ian Miell: That's right. Yeah. I didn't need to tell you to do that. I knew that's what you'd bring to the table.

Jamie Dobson: It comes naturally. It comes so naturally to me.

Ian Miell: Okay, so the thing that triggered the resurgence of this debate, recently, was a piece by David Heinemeier Hansson, I hope I'm pronouncing that correctly, who is CTO of 37signals. And he's talking about, in this piece that he wrote recently, he's talking about a company he, or a product, I'm not sure if it's a product or a separate company, but a company he runs called HEY. Which is some sort of email service, sort of productivity focused email service that was built a few years ago. And he's been blogging for the last six months, occasionally mentioning that they're planning to get out of the cloud. And this was the money shot really, this was the piece where he's saying it's a huge success. And this triggered a whole lot of discussion in the field. Basically, he claims that he's spending $600,000 on servers to save $7 million over five years.

And a lot of people critiqued it sort of in a knee-jerk way. They didn't actually read all of his posts. Because although it's quite a contentious post, or a provocative post, if you read his various posts on the subject and take away some of the hype, he's actually talking about quite a limited use case, and is aware of the arguments limitations. But many of the critiques of the piece focused on, he hasn't considered the cost of managing all this stuff off the cloud. He's naive about that. Or that he was saying everyone should get off the cloud now. Which, neither of those were actually fair critiques. And we'll go into that a bit more. But if you read this piece which is linked... Do we send the slides out, Carla?

Carla Gaggini: Yes, we do, in the follow-up email tomorrow.

Ian Miell: Okay, great. So you'll get the slides with the links. I put the links on the bottom right of the slides, if you're interested in looking further. But if you do go and read this piece, make sure to read his other previous pieces where he goes through his logic a little bit more, less sensationally. Anyway, yes, this is why the subject has come up again recently. And the debate on cloud repatriation is particularly hot, in the sense that there's a lot of vitriol on both sides. I've taken some of these out here. You can see Camille Fournier says, "I swear the idea that cloud workload repatriation, in quotes, is a good idea gets just more absurd the longer I think about it and it started out pretty high on the absurdity scale."

If you read down the thread, she gets also a little bit less attention grabbing in her statements. But it's not an uncommon response. Also, Corey Quinn, who's also very active on social media and works for a company that does cloud cost optimization, I think particularly for AWS, but not necessarily uniquely for AWS. In his "last week in AWS" blog, he talks about cloud repatriation and says it's not a thing. So absolutely, definitively, it's not a thing. Also, he casts aspersions on people who say it is a thing because he says that they have skin in the game, that they're biased because they're vendors selling colos, or they're vendors selling on-prem services in general. It's an interesting accusation to make when you, yourself, make money off the hyperscalers workloads. But we'll leave that aside for a moment.

And the absurdity of the debate, the extremity of the debate, sort of summed up while I was researching for my piece recently. I got this snippet on the bottom left there, where InfoWorld, in November, said "Don't buy the cloud repatriation clap trap," it's all nonsense. Then, literally two and a half months later, this is the year of cloud repatriation. So it's like Linux, this is the year of Linux on the desktop, you're going to get lots of noise for and against. And if you care to look at the blog post I wrote, it's really sort of a breakdown of these two very binary points of view into, hopefully, a more nuanced view. Which I'm going to going to try and communicate today.

Jamie Dobson: Ian, did we find any evidence for my grand hypotheses? Which was basically, that if you move to the cloud you realize straight away, bang, it's expensive. So you start to automate your workflows, you start to build a team that's really good at cloud and cloud engineering, and then when you've got everything containerized, automated and working properly, then oh, shit, you might as well bring it back in-house because everything's working, and you've got a great team and a great capability. And therefore, you get everything you wanted, which was a cloud native stack and no cloud bill. Is there anybody out there with this idea, except me?

Ian Miell: Yes, there is. Yes. And I'm not just saying that because you are my boss. Yeah, essentially. I mean, this covers a lot of things. Unpacking that statement is going to be part of this talk, really. But very briefly, if you're going to the cloud, there is an argument that you don't actually save on operational costs. Because the argument was that you go to the cloud and you don't need all those people fiddling with servers, and racking and un-racking, or dealing with colos or whatever. But one of the things that the HEY chap says, is that his operations team isn't going to grow by going back on-prem. That remains to be proved. But also, when he went to cloud, they didn't reduce their manpower either. And the other thing you touch on, which we'll cover later, is the innovation point. So is there a point where innovation stops being a priority and then cost optimization becomes a priority? So these are things that may mean that cloud repatriation becomes more of a thing. I'll definitely talk about that more at the end.

I mentioned that Corey Quinn accuses everyone of having skin in this game. And as Jamie's already touched on, we are a cloud native consultancy, so have some interest in this, but we don't particularly care if our customers run on-prem, colo, cloud, whatever. We don't care. We're entirely vendor neutral. We don't have partnerships with specific vendors in the space. We have been asked by some of our clients to help repatriate businesses. And the number of conversations we have about cost: isn't cloud expensive, or does it save us money, and so on. That's always a topic that comes up as we are consulting. I have to declare, and I suspect Jamie's of a similar mind, I have a slightly grumpy old man view of this as well, because often I look at cloud bills and wonder where all that money's going, when I could see how I could run a web server myself. Because I used to do that and run multiple for various businesses.

I cut my teeth working for a company that ran extremely busy systems on tin using only a few servers. And yes, there was a cost in terms of maintenance of that, but the cost of it was actually pretty negligible. And some of the cloud costs I see, I just keep thinking 20 years ago we could have run this with two people rather than spending half a million on it. So I have a slightly grumpy old man point of view here. There's a link there, also, to the blog I wrote recently on this subject.

Okay, so first of all, before we go on, let's define what cloud repatriation is. There's a slight misconception in some of the debates around this, which leads to strong views. Cloud repatriation doesn't necessarily mean we're going to move all our workloads off the cloud and onto servers under people's desks. That's a very extreme example of repatriation. So cloud repatriation is the process of bringing applications or workloads, either back from the cloud, or if you've never been on-prem or colo, trying the colo method of hosting.

The big poster child of repatriation is Dropbox, which many people will have heard about over the years. And I'll go into more detail about what Dropbox did. But even they still retained 10% of their storage capacity in the cloud. I don't know the specific technical details of what they did, but I imagine that was for bursting capability. But even they have not completely got rid of the cloud for where they've repatriated.

Most often, cloud repatriation is the movement of some workloads, or sets of workloads, back to some sort of rented or bought tin. And the more you look at this, the more complicated it gets, because there are cloud providers that will give you direct access to individual machines. I think Hetzner's the big one in this space, but there's Scaleway, maybe, as well. So if you want machines, you can hire them singly, or you can hire a section of a data center, or colo, or you can bring it back to a server room literally on your premises.

In my researches, I didn't find many examples of born in the cloud companies moving workloads to on-prem or colo. That doesn't mean it's not happening. Arguably, it should be happening to those companies more. And again, we'll go over that. But in terms of what we've seen, there's not many talking about born in the cloud companies moving to on-prem or colo.

Okay. Who's done it? When I started looking at this, I expected to find dozens of companies who've been blogging about what they've done in this area. To my surprise, I could only find three, or arguably two, or arguably even one, if you want to be really strict about it. As I said, the big one is Dropbox. There's also a mobile gaming company called Zynga, who were also well known in this area. And HEY, we've already talked about.

Dropbox had a project they called Magic Pocket. I don't know why it's called Magic Pocket. It sounds like there's a reason, but I don't know what it is. And essentially, Dropbox, we're probably all familiar with Dropbox as a product, and they're very heavily focused on storage, as you can imagine. And they use S3 to store files in, I think it was AWS. And their project was to emulate S3 on their own DCs. They essentially, I think it's petabytes or exabytes, I can't remember, some huge amount of storage they needed, and they were spending that in AWS, and presumably, also spending money on egress charges, and presumably, wanted to reduce the cost of that.

The project took them two years. I think finished in 2016, I think. And they produced an S1 statement in which they claimed that they had $75 million saved. I'm not an accountant, so I didn't comb through that in detail. But I think that $75 million saved doesn't include capital expenditure they made over those two years, which was less than $75. So the numbers banded around are about $53 million is what they spent on these Dcs, I think, in the first two years. So if you just take those at face value, that's $22 million saved. That's not to be sneezed at. And I also don't know how much those costs have continued over time, but we haven't heard of them going back significantly to the cloud. And if that did happen, you'd think there'd be a bit of a fuss about it. So seems that that probably worked for them financially.

Zynga, as I said, it's a mobile gaming company who used AWS right from the early days, and were a sort of poster child for AWS. Then, they decided in 2011 to build their own cloud called Z-Cloud. But five years later, they returned to AWS and published a case study in 2016. So this sort of can be seen as a repatriation failure, or perhaps they could see it as a learning exercise over those five years. People have pointed out that AWS massively increased over that time. So I think a lot of the scepticism around repatriation comes from that era where, perhaps, Zynger was sort of the original poster child for repatriation and then went back.

Jamie Dobson: Ian, do you think they went back with their tails tucked between their legs, begging to be allowed to return?

Ian Miell: Well, I speculate, and I have no hard evidence for this, but I speculate that they gave AWS a case study. There's a link to the case study in the slides. I would be surprised if they didn't get some sort of hefty discount for coming back.

Jamie Dobson: It's good that you speculate. We don't care about evidence around here. We care nothing for that.

Ian Miell: Yeah, people have told me privately that they've managed to squeeze discounts from some of the hyperscalers by threatening to move off the cloud. And you've got to assume that the hyperscalers are going to make a judgement about the seriousness of that threat.

Jamie Dobson: Yeah. We call that the Sky manoeuvre, now. That's going to be a new patent in the new book, the Sky Maneuver.

Ian Miell: Oh, is it? Okay.

Jamie Dobson: Yeah. You know Sky Sports? They charge you 120 pound per month for sports, and movies, and all that shit. If you ring them up and say, no, I'm going to leave now and just stream all this stuff, massive discount straightaway. Bang, the Sku manoeuvre.

Ian Miell: Yeah, I like that one.

Jamie Dobson: If anybody-

Ian Miell: If I want to change my mobile provider, yeah.

Jamie Dobson: Absolutely. If anybody's listening and you're still paying top dollar for Sky, you're foolish. Just make a call, tell them you've got financial hardship, basically because of what you have to pay them. Bang, gets fixed instantly.

Ian Miell: Tips from the CEO.

Jamie Dobson: Mate, you don't get to my position without thinking differently.

Ian Miell: Yes, CJ. Okay. Yeah. Zynga arguably turns the three into two. And finally, we've got Hey, who we've already mentioned, and they say they stand to save $7 million. They haven't finished it yet, but they've since produced a blog, linked down here, where they say it's going really well and better than they expected, and it's all lovely and wonderful. There's some interesting quotes from, HEY. The CTO says, that "Any mid-size SaaS business and above with stable workloads," and those are two very important statements, "that do not benchmark their rental bill for servers against buying their own boxes is committing financial malpractice." Now, that's a typically trenchant thing that that blogger says, but it actually gets right to the heart of it, really. Which is that once your workloads are stable, this may be something you want to consider for costs.

And also, he cite RDS and ElasticSearch as key products in their move. Those are two things that other people in the industry I've talked to have said are, sort of, eye wateringly expensive in the cloud. Especially if you know how to run them. So I've run DBs in PostgreSQL and Informix before. Informix particularly, oh, this is old man stories now, but Informix particularly, you could run at scale with one or two DBAs at most. And typically, you could get by with one, but you probably won't have redundancy in your DBAs if you're serious. The cost of hiring a DBA compared to the cost of running big iron databases in the cloud, probably a lot cheaper to hire the person.

And ElasticSearch is mainly about storing and egressing data. Managing data in the cloud can be very, very expensive, and ElasticSearch has come up a few times in my discussions with people in this area. But I thought this last quote was particularly interesting. "Anyone who thinks that running a major service like ours in the cloud is simple has clearly never tried. I've yet to hear of organisations at our scale being able to materially shrink their operations team, just because they moved to the cloud."

And while I was putting these slides together, it occurred to me that CloudOps can be a bit like government, in that government always says we're going to hire more people to do a function and eventually the taxpayer will save money. But of course those people never go away, and the numbers of government employees always go up. Is cloud management a bit like that? The promise never actually comes. So that's a very interesting angle to take on this and follow as we go through.

Jamie Dobson: I find that one. I'd love to hear what other people think. I find that highly, highly debatable. Not least, because how many of our customers have a had huge, huge, huge operations, started their journey to the cloud, and then their workforce changes, their capability changes, and they are indeed left with a much smaller traditional operations team and having their places, teams like SRE teams, or cross-functional teams, or whatever. So I'm kind of curious here, what's your take on that one?

Ian Miell: I've seen two things. One is operations teams don't significantly shrink, but usually we're dealing with those teams as they're investing. So it's natural that as you're investing in cloud migration, you're going to have more people helping you with that before it's stable, and then you can reduce. And we are not around, necessarily, to see how that goes sometimes.

The other thing I've seen is that... It is sort of an anecdote here from a previous company I worked for. They moved to the cloud, or they were doing a huge multi-year program to move workloads to the cloud. As part of that, a key service they were going to move to was RDS. We had this centralised team of 20 or so DBAs, who were managing data centres owned by this company, managing databases on those data centres. And those DBAs will be phoned up by one of the 6,000 development teams and say, "oh, we've got a query running slow, can you help us out? And they'd go, right, well, you need to put an index on that, or this isn't normalised in the right way," or whatever.

And when it was announced we were going to the cloud, everyone cheered. All the development teams cheered, "Yes, we don't have to wait six weeks for a server. We don't have to wait months for a database to be set up by this horrible operations team that we resent because they're always busy, too busy to help us." And these DBAs were, at first, quite fearful of their jobs. Then the phone started ringing. Dev team after the dev team were phoning up saying, "You will still be there to fix our database problems when we're on RDS, won't you?" And they'd say, "No, that was the whole point, you have to manage it yourself." And so these 20 DBAs were suddenly in massive demand from these 6,000 dev teams, because suddenly they realised the value of having these DBAs there. Moving to RDS does not solve your performance problems. It doesn't solve your database problems.

Jamie Dobson: Were the teams doing higher value work? If the teams stay the same size, do people do more high value work?

Ian Miell: Do you mean do they adapt to doing the DBA work themselves?

Jamie Dobson: Well, I suppose in the olden days, if you were just logging into machines and manually rebooting servers that had crashed, I guess a bunch of that's automated, then what would you do as an Ops engineer? You'd do something more valuable.

Ian Miell: Yeah, potentially. I think where the cloud shines is where you can automate upgrades. A lot of the times resiliency comes for free. Well, sorry, not for free. You pay through the nose for it, but it comes free in terms of effort. So yes, you could focus on other things. And the benefits of cloud are also clear when you can spin up database after database, or suddenly you want to do a parallel test of 10 databases at a time and they're huge, and it's business critical that you test your changes before they go out, or performance testing. You can do all that in the cloud with a click of a button. Doing that in a data centre can be impossible because you've run out of servers, or because you can't get hold of the central DBAs to facilitate that for you, or you need to get sign-offs, so on.

Some things are massively improved by using the cloud and some things are not. But anyway, back to the point, in aggregate, those skills were still needed. Those skills were still needed. And if they're not centralised, they've got to be spread across the business. And it's very hard to share someone across three teams who aren't in your line of business. Oh, another, these centralised DBAs were paid for by the central IT function, who were always resented as a cost centre. And they would've had to move to a completely different business unit, which meant that they wouldn't be shared with other business units, which would make the whole thing a lot more complicated. So anyway, the short answer to your question is, I think there would've been, in aggregate, more work for people to do.

Jamie Dobson: Yeah. There's lots of questions coming in, Ian, but I'm going to tie them up at the end.

Ian Miell: Okay. Well if they're relevant, bring them up because...

Jamie Dobson: I think this one is more of a general question. But just to let you know, lots of chatter, and we'll get you with the questions soon.

Ian Miell: Okay, very good. How do I go down? Oh, come on. Now I go down. Right. Let me see. Okay.

Jamie Dobson: I'm going to have a bit of a nightmare today, mate, with your screen sharing. You're not like an old grumpy dev, you're like somebody's dad. How does that? What do I do now?

Ian Miell: That fits.

Jamie Dobson: Yeah.

Ian Miell: Who's done cloud repatriation? It's quite hard to find names, quite hard to find big names who want to shout about it. And that can lead you to think that it's basically not a story, that it's not a thing. That these are the dogs that bark, but there are many dogs that don't bark out there. However, there's a really good piece by Andreessen Horowitz, or a16z, who is a venture capital company specialising in SaaS investments. And they put out a piece. It's quite interesting that they put out a piece as a VC, you'd think they'd be sort of more forward-thinking and more future focused, rather than saying get out the cloud. But they had this quote, which is very quotable. "You're crazy if you don't start in the cloud, and you're crazy if you stay on it."

And essentially, this piece, which is very well worth reading, makes the case that the benefits of cloud would really suit you when you're starting up. The elasticity, the flexibility, the possibility for fast innovation, the leanness with which you can grow is clear. So you've got to start in the cloud, pretty much, if you're a fledging startup that they're talking to. However, you reach a point where the cost of these things outweigh the benefits, and then you're crazy to stay on it. And this chimes with what the HEY CTO says, "Any mid-sized SaaS business and above with stable workloads that does not..." Well, I quoted this already... "does not benchmark their rental bill for servers can be accused of committing financial malpractice."

So if it's so crazy to stay on it, why are we not hearing more names? And again, this is speculation. In my researches, I found consultants who specialise in this area were talking about some of their clients who had reason to move off the cloud. There's a screenshot of Dell, Michael Dell, himself, boasting about bringing someone back into a colo and saving the millions on their cloud bills. People who do this aren't really going to talk about it. It doesn't seem to be something that happens as a global strategy across the business. It seems to be something that happens more piecemeal for particular business units or in particular use cases.

Jamie Dobson: There seems to be just ad hoc cases, don't there Ian.

Ian Miell: Yeah.

Jamie Dobson: I remember when I worked for KLM back in the day, they had their own data centre deep, buried underground with its own backup power generators, because obviously their systems could never go down. But how many people really need to go off grid? Hardly any. Maybe it's the same here. How many companies really need to not be on the cloud? So maybe this is why we see ad hoc cases. Are they similar in any way or are all these cases different? So we know there's specific use cases.

Ian Miell: I'll get to this in a bit. But to summarise now, generally, the big driver is cost. So maybe a new CTO comes in, or CIO comes in, and new broom can chuck away old stuff. So you get cases where they try and move some workloads back to the cloud to see what the cost effect of that is.

Jamie Dobson: Now, here's an interesting thought from the comments here, or the questions here in the chat, Ian, that I think add something to this a bit. So this person, John. Hi John, how are you doing? So John says, we are shifting some of our AWS managed services to be RO managed services, such as MQ broker. The reason is to be more in control, and sometimes AWS limits what we're doing. So in other words, cloud has made the easy things easy, but the hard things harder.

Ian Miell: In a former life, I was working on cloud solutions for Kubernetes. And yes, the fact that you could spin up a new cloud instance with a click of a button was great. But when dev teams, one of these 6,000 dev teams came to us and said, we want this specific thing running on EKS, sometimes it couldn't be done, or it couldn't be done easily, or it would take a long time to deliver because of the nature of where we were. So in certain respects, yes it can. If you're running open source software on your own servers, as many people here will be doing at home, you can do what you like. You can recompile stuff, rewrite stuff, you can try a new version, you can apply a patch, whatever you like, but then you're responsible for your own mess. So there are those swings and roundabouts.

But yeah, it's interesting. I would've thought... It's interesting that MQ broker users are talking about control over cost, before cost. I imagine they're both mixed in there somewhere. Maybe we can ask the person is cost a factor for them, or is it really about control? But generally, the more specific your needs, the less likely a commodity product is going to help you. So when I was working for, I may have mentioned before, I worked for a company for 15 years that was late to VMs, didn't use VMs until about three, four years ago. They stayed in on-prem because milliseconds counted. And in the early days of VMs there were real problems with reliability, in terms of if you do something, how sure are you going to be that it is going to complete within two milliseconds or 30 microseconds? When these things matter, then that control makes a difference.

Okay, let's move on to the next one. We're not seeing many examples of people shouting about repatriation. In my researches, I found other surveys that said quite eye grabbing things, like 48% of decision makers migrated apps or workloads from hyperscale providers. Which sort of made me furrow my brow, because I thought that's a lot, surely we'd have seen that in the cloud revenue numbers. But the closer you look to these surveys, as Corey Quinn says, you have to be sceptical. And these particular users were 600 IT decision makers who were in colos. So they went and asked a bunch of people who are already in colos, "Have you done repatriation?" And of course many of them said yes. So this doesn't really reflect, necessarily, the numbers in the industry. Similarly, OpenStack trumpeted that the OpenStack cloud has grown by 166% since 2020. That was published last year.

And again, I don't have access to the whole details. And again, I'm speculating. But the fact that they promote number of CPUs... So when they're talking about 166% growth, they're talking about number of CPUs running on OpenStack. Suggests to me that they're not seeing explosive growth in customers, but the customers that are using OpenStack have grown their estate. So you have to be really careful about numbers that you see, and what they mean and what they don't mean, and what they say and what they don't say. But as I mentioned before, you can find examples of videos on YouTube where consultants are talking about, not directly who they're talking to, but some of the conversations they're having. I've linked to some of them there.

In summary, it is a thing. But how big a thing it is, it's very hard to determine. There's also a trend for cloud refuseniks, people who say, "We've never been in the cloud and we're not going to cloud. So nah, it's never going to happen and we love it." So again, there's a survey that said 98% of businesses are on-premise. That's another survey that's got to be questioned because it's from Spiceworks. And it's with Spiceworks own customers, who tend to be small businesses. Small businesses probably don't need the cloud, they don't have enough spend for the cloud, and so on. So that 98%, again, take it with a huge pinch of salt.

Jamie Dobson: We have a prediction here, Ian, 2024 will be the year of OpenStack. What do you think?

Ian Miell: I don't know, I'm not seeing evidence of that.

Jamie Dobson: It might be.

Ian Miell: Are they winding me up?

Jamie Dobson: Somebody might be winding you up, yeah.

Ian Miell: Yeah.

Jamie Dobson: We do have good questions coming in though, Ian, but I think these are the ones that I'm going to save to the end.

Ian Miell: Okay, very good. And recently, there was another sort of similar article to the HEY one, how Ahrefs saved $400 million in three years by not going to the cloud. So there's sort of people starting to come out the woodwork and say, hey, it's trendy now not to be on the cloud. You can get some good attention by doing that. So we're going to talk about why we're doing it. Another trend, which I don't know very much about, but it's definitely in the air, it's vertically integrated cloud hardwares. There are companies that are now specialising, their sales pitch is basically that they can give you hardware, which is designed to run cloud type workloads and has a software on it that can do the whole thing for you. So if you like hardware and software, OpenStack type things. Oxide Computers is one of the famous ones. There's someone very, I can't remember the... There's a couple of famous sort of, not household names, but tech household names.

Jamie Dobson: Oxide is just-

Will McDonald: It's Bryan Cantrill.

Jamie Dobson: Bryan Cantrill.

Ian Miell: Bryan Cantrill, that's right. Yeah. And Jessie Frazelle.

Will McDonald: Who, himself, is quite opinionated. Yeah.

Ian Miell: He's fairly opinionated, yeah.

Jamie Dobson: Bryan's a very, very good public speaker and he learned a lot from me when we shared a podium once. That's all I'm going to say there. Oxide is very esoteric, says Hugo. This is correct, Hugo, it's very esoteric. They've been quiet, Oxide, actually, they've been quite quiet for a couple of years, at least in my Twitter stream.

Ian Miell: It's very hard to tell. I mean, they raised money.

Jamie Dobson: But even we did that.

Ian Miell: Sorry?

Jamie Dobson: Even we did that.

Ian Miell: Yeah, even we did that. But they raised money, which doesn't tell you anything. But they're still around so maybe they've got some big customers, maybe they haven't. I don't know. Even if they had customers, they're probably probably not customers who want to shout about it either. So it's pretty hard to tell how big this is. But SoftIron's come up a few times in conversations recently in this area.

So the other thing about this is, to me, this sounds like mainframes. And there is an argument why they're not mainframes, but I don't really understand it. I did look at it on Hacker News and there was a long debate about it. But essentially, if you're paying for a big machine that can run a load of stuff in a vertically integrated way, sounds a bit mainframey to me. But presumably, it's mainframes that can run the latest stuff in a more native way. I don't don't know exactly. Or perhaps you don't have to pay eight figures every year to IBM to have them either. But if anyone's got any intel on this, real intel on this, I'd be really interested to know.

Jamie Dobson: Oh, here's a great question from my friend, Willem. Isn't the cloud just a big mainframe?

Ian Miell: Yeah, huge one.

Jamie Dobson: Yeah, it's a huge one. But it's more like a mainframe with ASR33 Teletype machines connected to it. It's a time-sharing computer with a slightly better interface, but only just.

Ian Miell: You're not old enough to remember teletype, are you, Jamie?

Jamie Dobson: I read books.

Ian Miell: I'm not even old enough to remember teletype.

Jamie Dobson: No, but there's a few people from the sales team. Just kidding, guys. Just kidding.

Ian Miell: Okay. Why would people refuse to use the cloud? Well, this is related to a lot of the reasons why people would use cloud and why they'd come off it. So cost. Again, we've seen with Ahrefs' performance, we've discussed already. Security and data sovereignty is another one. Security is one that gets people's goats often, if you talk about, I'm not going to the cloud because it's not secure. And then they'll throw lots of ISOs at you and reasons of why the cloud is secure.

But while the cloud actually has a genuinely good reputation for being secure, there's two reasons people might be scared of it, CIOs might be scared of it. One, is that it's quite embarrassing if you have a security breach and you're using public cloud, because those breaches can be significant. So the famous one was Capital One, I think they had S3 buckets exposed to the public. And I can imagine that some board members might have turned to the CIO and said, why did give our stuff away to someone else? And they wouldn't have necessarily understood the nuances, the fact that an engineer didn't click the right button when creating their S3 buckets.

And the other reason, is that there may be regulatory reasons for you to stay, to require that you have the sovereignty over your data more under your control than the cloud can give you.

Jamie Dobson: We've got a fascinating comment in the chat here, not just about data sovereignty but hardware sovereignty. So SoftIron make a big deal apparently, according to Charles, about their hardware prominence. So the only thing they don't make is the CPUs, but everything is assembled and built from American components, i.e., not Chinese components. And that plays very well in the eyes of the US government and military, et cetera.

Ian Miell: Right. And we've got Gov Cloud, so there's whole clouds, or data centres and regions, devoted to government use alone, presumably because they don't want their stuff to be run alongside other stuff, with any question that it's going to be run anywhere near anyone else's stuff. Oh, Jamie, you're on mute.

Jamie Dobson: Well spotted. I was going to say, again, these are very specific use cases, aren't they? The US military, et cetera, et cetera.

Ian Miell: Very specific.

Jamie Dobson: Yeah. Most people wouldn't have such constraints, would they?

Ian Miell: No, the vast majority wouldn't. No. Absolutely.

Jamie Dobson: Just a quick shout out on this moment, just a tiny shout-out going back to punch cards and teletypes. Steven Porter, welcome. Steven claims here that his first terminal was a teletype and his first job was punch cards and paper tape. So welcome, Steven. I'm not sure anybody has ever made that claim on one of our webinars before, but how fascinating.

Ian Miell: Did you know my dad used punch cards? So my dad's a Luddite, but when I was born, on my birth certificate it says, father: computer programmer, in the job. I spotted this five years ago and I said, "Dad, what the hell is this?" He said, "Oh, yeah, when you were born I had to take in any job I could get. And basically, the machine would clam up and there'd be some blockage, and we'd kick it and go for a pint." And that was his job as a computer programmer then. So nothing's changed, really.

Jamie Dobson: I'll give you one more bit of trivia, historic trivia, and then I'll let you get back on with your talk. The reason we call compilers compilers is because we'd use punch cards that has rotuines we would use multiple times. So you'd grab a few punch cards that were in your library and you'd make a pile of them to feed into the machine. So the verb, to compile, came from putting the punch cards together in a pile. And then later that was automated to become compilers. There you go.

Ian Miell: I can tell what you've been reading recently, Jamie.

Jamie Dobson: It's that obvious.

Ian Miell: So yeah, security data sovereignty. And the other one is just there's no obvious business need to move. A lot of the reasons people go to cloud is from pressure from their own staff. So anyone old enough to remember when Java came out, every developer in the world was like, "Port out code to Java, or write new applications in Java, or I'm leaving." And a lot of companies got scared of that and decided to go to Java. And there wasn't any obvious need for them to go to Java, it was just` if you wanted to retain staff, you do that. There could be a similar need with cloud. If you don't have those engineers saying those things, then maybe there's no obvious need for you to go to the cloud.

Who should not repatriate? We've touched on these things, but it's good to have a list of them. So, people who should consider repatriation. Pattern seems to be, if you've got homogenous workloads, so Dropbox, they're just storage, they do one thing over and over and over again. So if you've got homogenous workloads, HEY is the same, it's basically a three-tier app, monolithic database, so on. These kind of workloads don't necessarily benefit from the cloud. They're not cloud native. Well, S3 is cloud native, obviously. But if you've got these homogenous sets of workloads, then you might want to consider repatriation. If you've got a stable product and no longer innovating... So talking to people in the industry, one of the things they said about using the cloud was that it allowed them to experiment very quickly with improvements to the system.

The example that was most vivid to me was someone who was a CTO of a company that ran very variable workloads, and they wanted to determine where their bottlenecks were. Basically, they just one day decided to use very fast SSDs to see what effect that would have on the system. And within a day or two, they had the answers. They found that it made things go much faster, and also, they discovered a bottleneck later in the chain. Now, doing that on data centers is, potentially, it's going to take months for you to procure the hardware, sign-off the hardware, deliver it to the data center, plug it in, move the data, and so on. So the fact that they were in the cloud meant that they could do that really fast.

But if you're no longer innovating, if your workloads have settled down into a steady state, then the time may have come to think about repatriation. Cost, we've talked about, and it can be a clear business need. The expertise available to manage. So one of the things that inhibits many companies from moving back is the fact that no one knows how to do this stuff anymore in your business, or hiring people who do know how to manage web service by hand is becoming increasingly rare.

And there's actually a thesis mentioned on a Hacker News post I'll reference later, which says that's the main reason it's not even more significant than it is at the moment. If you've got a clear business need, like a regulatory challenge or something like that, then that's another reason. All this adds up to... The Andreessen Horowitz piece talks about SaaS companies. SaaS companies are a sweet spot for this. They needed innovation, they needed the ability to move fast, they needed the ability to scale fast, so they went into the cloud. Then it reaches a point where they're in cloud, the cloud costs are significantly higher than their dev costs, and so that becomes the target. So here, people who shouldn't consider repatriation: wild swings in usage, need to innovate quickly, don't care so much about cost at the moment.

Jamie Dobson: But the SaaS thing, Ian, is not just a mature SaaS product, but it's a brand new SaaS product. So in the beginning of your cycle, you spend way more on marketing, and it's all about users, users, users, much less about revenue. That might have changed, by the way. So all of a sudden investors, like Horowitz, Andreessen, all of a sudden they care about profit. And that's a multiplier. But it used to be users and top line, even if you were wasting money. So the cloud let you scale really quickly. That's a specific case, and as you said, once it's mature... But if you had a mature size product that was on-prem, would you then think, okay, I'll go to the cloud, sort it all out and then come back? Probably not.

Ian Miell: Probably not, no. No. Well, that brings us on neatly, Jamie, because I'm just going to cover something you touched on there. What's next? It may be topic of the week, but there's more and more people posting about this kind of thing. We saw Ahrefs got in on the debate. There was a recent post from Michael Dell boasting about using Dell hardware to repatriate their AI analytics workloads. So ticking all the buzzword boxes there, from a hyperscaler to a co-location provider. By the way, very few people talk about going back on-prem. So no one seems to want to go back on-prem, but a lot of people talk about colos. It's almost like it's a purchasing sweet spot. You can pay for people to manage the hardware, rack and stack, plug cloud cables in. That's that's a cost you can bear. But the hyperscalers value add is not necessarily what you're after.

Jamie Dobson: So there's two problems though, isn't it, with Michael Dell. Clearly he's got a horse in the race, he sells servers which will be part of a colo hybrid solution, right?

Ian Miell: Yeah. You're not going to find many people without a horse in the race.

Ian Miell: But yeah, as I mentioned earlier, 37signals are continuing to post about, they're doubling down on this message that it's saving them money and it's working for them. So it sounds like it's continuing. And this leads me to speculate, is there a macroeconomic reason why this is going on? If you look at the history of interest rates, you can almost draw a line there, where interest rates are low, and they were low until beginning of, well they were low until this year basically, or the end of last year. Where that line is low, you tend to get more investment and innovation. Money is flying around because money is cheap. And as money has got more expensive, you got to wonder whether more and more businesses are going to start looking at their costs. And in fact, the 37signals guy says, don't reduce your payroll, because there's been a lot of layoffs recently in tech, don't reduce your payroll or reduce your cloud cost.

So it seems that people are starting to look at this. And I'm not an accountant, as I mentioned, but in times of plentiful money, you get a lot of potentially OpEx spending going on. But it may be that the accountants may come back and say, hey, if we do some CapEx, we can depreciate it over several years. Hardware seems to last longer. That's something else. People are saying that the typical shelf life of hardware used to be three to five years, and now seven is normal. And many say, well, we use it far beyond seven. It'll be interesting to see if interest rates continue to go up, will there be an increasing emphasis on cost-cutting, and therefore, many more businesses are going to look at this option.

Also, this is taken from There seems to be a slowing of cloud spending. And similar to the OpenStack 166% argument, this article was interesting because the growth has slowed down. So we're not talking about deflation or reduction in cloud use, we're talking about reduction in growth. However, it's gone from 33% in Q2 last year to 20%. And buried in the article is also mentioned that, in January, AWS revenue growth was in the mid-teens. So this may explain why AWS has suddenly started laying off, not huge numbers of people, we don't know the actual numbers, 9,000 was the latest round, but we don't know how many people work for AWS. It's quite a lot, it's in the hundreds of thousands. But anyway, you can see why they'd be suddenly a bit nervous. And I'd be interested to know how many new customers AWS are getting at the moment and how much this is existing customers increasing their spend.

Jamie Dobson: I think the economic pressures are definitely a thing. Every time there's changes, whether it's inflation or recessionary pressures, it's always about cost-cutting. And even if you just pause. So I think there probably is a relationship between... Maybe it won't lead to repatriation, but it might get everybody talking about it. I definitely could imagine, if you're humming and hawing about a cloud move, you know it's expensive to do it, recession arrives, everybody pauses, Amazon's shitting their pants, lay people off. Everybody holds their breath.

Ian Miell: Yeah. Interestingly, they say it's going to continue for another two quarters, they're projecting. That's what their CFO is saying. And finally, I found this this morning, it was a quote from Hacker News, where I think it's a professor says he had a student who did a thesis on de-clouding, which was apparently the academic term for cloud repatriation. But I underlined some interesting things.

"As companies move their operations to the cloud it's not the disappearance of hardware on-premises but the loss of skill-sets. Later they don't even know how to write a job description to bring people back in." Which, to an old person like me, is quite surprising. But try finding a CIS admin who can do simple tasks, like set up and properly configure a mail server. That person has never set up a mail server. I can tell you that now. No one thinks that's simple. But people who could have a go at it seem to be rarer.

Jamie Dobson: I tried to do it once. I've got a vague memory when I was about 21, trying to do all this. I was like, what the fuck is this?

Ian Miell: Yeah. I remember the SMTP book that was on some people's desks, was about one and a half thousand pages long.

Jamie Dobson: That being said, the best thing I did is I once wrote a web server. Because I didn't understand HTTP protocols and I wanted to understand it. So I wrote a website. But that was really cool. But then I can't be bothered to add HTTPS.

Yeah, I miss those days. I definitely, Ian, I think the cloud providers are banking on this. They're banking on this. At the turn of the last century, okay, this is my last historical point. Turn of the last century, when the electricity grids came online, obviously over 20, 30, 40 years, all previous electrical engineers, who used to work in factories, no longer worked there. There was, instead, a much smaller workforce running the national grid. Surely this is part of the cloud's long term play. Everybody will de-skill, and there'll just be a handful of people within the cloud providers who know how to do this stuff. And then you can't repatriate, even if you want to.

Ian Miell: Yeah, absolutely. It still amazes me that Amazon made a huge amount of money by just provisioning web servers.

Jamie Dobson: Yeah. Giving you SMTP starting now, so we don't have to.

Ian Miell: Yeah. There is that argument that they're investing in that cratering of skills. And from a selfish point of view, if those skills come back in fashion, then that might be my retirement sorted.

Jamie Dobson: Oh, sign me up for a bit of SSH. I miss all that stuff. I don't think I could do it anymore.

Will McDonald: We'll be just like COBOL developers for Y2K.

Ian Miell: Yeah.

Jamie Dobson: I'm old enough to know Java when it was trendy. It was very trendy once upon a time. Now, Ian, I think I speak for everybody on the call, we're very grateful for all the research you've done and then pulled together into a really cool, cohesive narrative. Are you now ready for questions?

Ian Miell: I am.

Jamie Dobson: By the way, before we do that, I want to give a shout-out to Andrew Moloney. We mentioned earlier, SoftIron, the company. Well the chief marketing officer is on the call today. So, welcome. Your product got a shout-out and we didn't even know you were here. Now I know what Andrew's thinking. How can I get involved in sponsoring events like this? So I'll just drop the link in the chat box there for anybody. Something you can look into for our next conference. But it's good to have a cross section of the community here. So welcome, Andrew. Right. Questions? Ian, we've got a few questions. First one, is it an intelligent decision to repatriate? Probably not, by the way. Or is it better to understand more about cloud strategy and continue there?

Ian Miell: Well, it depends, right? I think Dropbox would say it is a good idea for them because of their specific circumstances. So because they had this very constrained load, a mature product, spending absolute tons in the cloud, millions, and millions, and millions. If you have to go out and pay half a million dollars a year for some expert in something esoteric hardware related, then you can do that and it can be justified. But for many businesses, I think, especially if you're using lots of different bits of the cloud, unpicking that is going to be really, really hard. So I'd have to explore the specifics of those that use case to know whether it's a good idea.

Jamie Dobson: I've got a question that's come up from today, Ian. I don't know if we can answer this. And I say we because we work on the same team. The person wants to know the top five reasons why we would advise a client not to go to the cloud. Now the reason I said this might be hard to find five reasons, is because if a customer, for example, is off the grid, we're doing some air gaps cloud work, then obviously they know already that they're not going to a public cloud or air gap Kubernetes. So what would be our top five reasons, mate, for a customer not to go to the cloud?

Ian Miell: I can think of one very specific example. I'm obviously not going to name the client, but we went to one client that was a very big company, and they had a system that was basically their cash cow. It was written in a fork of a language of the COBOL era. I can't remember what it was. They had to hire three people to maintain it. And they were saying, "We've got to go to the cloud because this is so old." And we kind of said, whoa there. First of all, it's going to be really hard to move that to the cloud because you've got 50 years worth of business logic buried in there. You can go to the cloud, start with a mobile app, put that in the cloud, then do your main website. Build the expertise and the muscle memory first, before you get your crown jewels and try and move that to the cloud, because that monolithic kind of big migration, it's going to be tough to make that work.

Jamie Dobson: I remember visiting a finance customer, but it was problematic. Everything was running COBOL from 50 years ago, but at the same time, there was nobody left who could edit COBOL. So the solution was, can we start to dismantle this? But it was extremely difficult. Or do you have to then retrain... One of the strategic solutions to this was retrain graduates in COBOL and pay them a massive amount of money to run this. So in the one hand, it was like, okay, maybe we don't need to think about the cloud now, but how can we get this whole stack of, essentially APIs, from the COBOL mainframes into something more modern and manageable, and then maybe later go to the cloud.

I know another reason is, not now. So some people come to us, we want to go to the cloud. And it's crazy. For example, there's no idea of the new roles that will be needed within the business to support the cloud, there's no re-engineering done. If you pick up stuff and it's just start moving it to say, okay, now we're on the cloud, a quick ramp-up will lead to a very embarrassing and quick ramp-down. So we might not say never, but we'll be like, you need to slow down, you need to think about what you're going to do. You need to do a cloud native assessment from Container Solutions. Right? And when you understand them and you de-risk, then you can think about the public cloud. So it's not a hard no, but it's also not a hard yes.

Ian Miell: Yeah, lack of skills, the lack of the muscle memory I talked about before. If you don't have that, you might land on your face. The other thing, another reason, is there can be naivety. People think it's going to be cheaper just because you think you're going to be able to stack a bunch of people, and therefore it's going to be cheaper. And that can be a naive way of looking at it.

Jamie Dobson: There's also, do you remember outside here on the office on Cannon Street? There was a guy, we told that guy about what we did, and he said, should I move to the cloud? And we were like, no, because you sell hotdogs off the street. That's a specific use case where we felt the cloud couldn't help him. No discount for the free consultancy.

Ian Miell: Yeah.

Jamie Dobson: Unbelievable. Here's a good question. I'd like to answer this, but it's your webinar, mate. Is cloud repatriation the trend? I think we probably, no, is the answer. But should vendors be worried and start reviewing their pricing model?

Ian Miell: Well, if you are a CIO who's spending a lot of money in the cloud, you could leverage this moment. The dark arts of CIO-ness involve, often, strategic plays, like threatening to move or starting an initiative to move off the cloud, and then letting them know about it in order to negotiate a discount. So yes, we may see that. And I know a few people have come to me and said they've managed to get discounts off of hyperscalers, if they're big enough to do that.

Jamie Dobson: If you go back to slide 33, my answer to the question is yes, they should be shitting themselves. And it's more or less, I think, because cloud will become commodity one day. Currently, we're in a trough. Everybody went to the cloud and now Amazon are firing people because everything's slowing down. But the ultimate destination for cloud will surely be pure commodity, and then it's a race to the bottom. That's good for consumers, but is it good for them? That being said, cloud will probably change into ChatGPT at high level services. I get that. But I think they should be worried. I think they should be worried, where does the next 10 years worth of revenue come from.

Ian Miell: Well, also, if it becomes the norm not to go to the cloud.

Jamie Dobson: Then they definitely should be worried.

Ian Miell: Yeah, because there was a time when, if you're a CIO and you weren't in the cloud, you'd look like the numpty of the group. But now, you might look like the hero. So these fashions might have a material effect on these businesses' ability to get revenue.

Jamie Dobson: I'm going to pick a couple more questions, Ian. Some of them have been answered in the webinar, so I'm not going to ask you them again. But this is an interesting one. Repatriation sounds like the sort of thing lots of people want to do, to lower costs, et cetera, et cetera, but few people actually do it. Why do you think that is, and what are those barriers that are stopping people optimising their infrastructure spend?

Ian Miell: As a former developer who thought we could do everything in a weekend, I suspect a lot of people are doing little proofs of concepts in their spare time. But scaling that up to actually a real effort to move stuff from one hosting environment to another, it's pretty daunting. I suspect a lot of those people struggle to get the business place clear enough to actually invest to do it, and they think they can do it in a weekend or prove that it can be done in a weekend, and that's job done. But I think it's building the business case and getting the ears of the right people. So if the CIO is not interested in repatriation yet, they're probably not going to be responsive to that kind of discussion. So that's what I reckon. I reckon that's one of the reasons it's not happening more. But the thing is, we still don't know how much it is happening and it's not being discussed openly. Or it's not a strategy that people are consciously doing, it's just happening sort of accidentally.

Jamie Dobson: Maybe the time's not quite right. People are humming and hawing, maybe in a few years as the technology gets more mature or the landscape changes, et cetera, et cetera. Final question. Final question, and then I'm going to open up the mic to anybody who wants to grab the mic and shout out. How do you decide on workload placement when it comes to cloud repatriation, Ian?

Ian Miell: How do you decide on workload placement? Do you mean what platform you should use?

Jamie Dobson: No, what are the key factors to determine whether or not we can move workloads back on-prem, or is it better to stay in the cloud? I think you answered that, didn't you? Around the stable stack, the less innovation, SaaS products that will mature.

Ian Miell: Exactly. Yeah. It's a range of factors, together, to make the case compelling or not.

Jamie Dobson: On that note, I'm going to open up the floor. Does anybody want to grab the mic? You are allowed to say thank you. Ian likes that. He'll put that in his next performance with you when we speak. However, you're also allowed to ask a question or state a bold opinion. Does anybody want to grab the microphone, or does anybody have anything to say?

Ian Miell: I'd be particularly interested if anyone's actually done this.

Jamie Dobson: Yes, of course. Has anybody done it?

Ian Miell: Or decided not to, or thought about it. That would be really interesting to me.

Everyone's being very shy.

Jamie Dobson: The problem is, with webinars, as soon as you get over seven, nobody wants to speak. Which is fair enough, which can be a bit intimating in a large group.

Willem-Alexander Dous: I'll go, then. At the very, very small scale, so we're talking about the, I think, can be expensive for small things. So for a little product I made for a few people, I chose a backend as a service with Google, with the idea I don't have to make my own login stuff. And there was a remark about if you outsource security, but you're still responsible for it. So last month, a user found a problem with their authentication. Six weeks ago. Yesterday, I got a notification from Google, look, there is this problem with our authentication. And it's been there, if you Google, it's been there for a year. So next small thing, it's probably going to be on DigitalOcean or Hetzner, unless it's a perfect fit for AWS Lambda or something. You've got a million sites apparently, running on this thing, and you can't manage the roadmap. So next small thing is moving back, but I'm not seeing that with big clients yet.

Ian Miell: That speaks to the control thing. The thing is, you can blame Google. If AWS has an outage, you can blame AWS. And everyone else is in the same boat. So again, this is...

Willem-Alexander Dous: That was the reason. But when people can't log into your application for a month, then you don't have product. So yeah, that's a good reason to do that. That's why I went there in the first place.

Ian Miell: Thank you, Willem.

Andrew Moloney: Hey, it's Andrew here from SoftIron. To give you our view, in terms of what we are seeing, I can't think of any project right now that we're involved in where it actually is a repatriation project. It's actually, they're all new projects where they're trying to build that kind of cloud experience internally rather than just repatriate workloads. I'd say that's what we're seeing. My guess, well, my hunch is, if it's going to happen, it's probably going to happen now. We're kind of at the beginning of that process, and I wrote a post around that, that maybe it's combination of the economic climate, IT skills, the cost of energy. All of those things, together, might be the kind of forcing function that starts that kind of repatriation happening. But I wouldn't say right now. We're seeing hardly any of that. That's all new projects and creating kind of cloud native on-prem or in colo.

Ian Miell: That's really interesting, Andrew. Thank you for that insight. That's not entirely what I was expecting, but that's really, really interesting. Thank you.

Jamie Dobson: Basically, you've got something and you think we need to modernise it and sort it out. Well, hold on. Maybe we don't need to go to the cloud, but we can do something different.

Andrew Moloney: So I want to replicate the experience, but I don't want to actually put it in public cloud.

Jamie Dobson: Our time, unfortunately, people, is more or less up. So I am going to use this to make a few further announcements. So first of all, this is being recorded. I forgot to mention at the top of the meeting, if you don't want to appear in the video, let us know and we'll pixelate you out. That was number one. The second thing is, there will be a feedback form. Ian and I have been threatening to have a more conversational webinar, like this format today. It's slightly different to what we usually do. Give us feedback. If you thought it was fun, entertaining, useful, let's hear it. But more importantly, if you thought I needed to speak less or speak more, or the flow was not quite right, give us a critical feedback and we'll work on this. I promise. If, in general, you've got feedback about Container Solutions, WTF, or anything like that, we are always listening, send it to us and we'll incorporate your feedback into what comes next.

Second thing, we are a community. We don't function without you. So it's nice. We had a few familiar faces in January, which was great. But since then, we've had both familiar faces and new faces, so that's really good. Please spread the word. Invite your friends. Come to the conference if you're in London, or if you're going to be around. The more people who come into this thing, the better. I now need to do my round of thank-yous.

So first of all, my big thank you goes to Carla. Without these things happening, they just don't happen. So round of applause to Carla.

Carla Gaggini: Thank you.

Jamie Dobson: Very, very special round of applause to me for facilitating and coming in and introducing Ian. But we will reserve the biggest round applause to Ian, himself, who's done a brilliant, brilliant webinar today. And he's done all the research that none of us could be bothered to do. So thank you, Ian. It was a really, really good one today. Cheers, bud.

Ian Miell: Thanks, Jamie.

Carla Gaggini: Thank you.

Ian Miell: Thank you, everyone.

Jamie Dobson: And on that note, this webinar is over. We wish you wonderful Friday's, accept it's only Thursday, but we're nearly there. We're nearly there. And after that, a wonderful weekend. And we will see you all next time. Bye Bye.

Carla Gaggini: Thank you, all. Bye Bye.

Ian Miell: Bye, Everyone. Thanks for coming.

Jamie Dobson: Bye.

New call-to-action

Leave your Comment