 Fygoedd i chi, gallu gyda'n gwylwg. Fel ydych chi'n ddweud ddim i gael i chi. Felly, rydyn ni'n Andrew Agabac. Fygoedd i'n llai ar y Cyfrofod Bywyd Agaibach. Fyrddwn i'w parwyddo, a gwylwch chi'n mynd i ymweld. Dwi'n mynd i'n hath o wath ddiolch, ond yn gyfraithio ar gyfer hynny o argymau gwylwg a'r cyffredig i gyfraith. ...egwyd yn mesafol iawn iddyn nhw'n golygu'r cyfeiri ar y wyernessol. Goeddwyd wedi bod gennym i chi'n gweithio fy hunain y fainfodd, o fawr o gyfoes, o'r cyflei ar wgweithio, a yn amlwg o'r lleachau sy'n gwybod hi ar y cyfansbeth... .. ac mae gennym ni'n mwy o'r gweithwyr yn yn ôl... ..a llwygarol yn gwybod yn cyflots i'r gwybod hyd yn ôl... ..y bobl fflau wedi'u psychnol. ac yn y gwybod y maen nhw'n helpu a bod yw'r ysgrifennu hynny sydd wedi'i helio i helpu, ac uchyddiadum i'r rhan o'ch llenn politically i ddweud y rhan o'r sgwrs. A o beth, ychydig gyda'r 1 yr lyfyn ysgrifennu... Ynchy mewn ysgrifennu, rwy'n cyfrifiad sydd wedi ei hyn i llunio ymlaen... phobl o rhan o'r cyfrifiad ymloedd ar y cyflugiau cyng각. Rwy'n cyfeirio gydag yma, yfynig yn 100 o shaped bod yw ysgrifennu, a swydd ni'n wnaeth y bydweud ar y dylau ymglygau cyfeinidol iawn, ar y bydd ymgylchedd arni. Mae'n rhaid i'n falch chwell a chwell ei bobl yn rhaid fyddwn ar gyfan cyfeinidol iawn. Rydym yn amlwgol iawn, a rhaid i'n defnyddio hynny, ac y mwyn ffant, ar ddiolch a fydd yn gwneud hynny, Rydyn ni wedi'i cyd-dweud ar y cyfnodd ymlaen o'u ffordd i fod yn ymlaen, er mwyn ymlaen i'r ffordd i gael ysgol. Ymlaen i'r ffordd, mae'n amlwg ymlaen yn gyffredigol, ond mae'r ymlaen yn ymlaen o'r ymlaen, yn ymlaen i'r ymlaen. Ymlaen i'r ymlaen o'r ymlaen o'r dwylo'n mynd i ddim ymlaen. Mae'r ymlaen i'r ymlaen o'r ymlaen o'r ymlaen i gael ymlaen is not so much run cost reduction, but actually the agility and efficiency in making change go faster and more effectively. Secondly, that sort of lift and shift your existing servers into infrastructure as a service. Yeah, there's some benefits there, but the real prize is actually when you move up the stack to platform as a service and software as a service. The third conviction is that in addition to all this conversations about the platform and the technology, actually the ways of working, the engineering processes, the software development life cycle, changing that part of how you work is just as important, and I think there's some cool connections with the way the open source community thinks about writing software in extended communities and how change happens there. And then maybe this is controversial, so feel free to disagree, ask questions and shoot me down, but I think my view is it's much more valuable to go with a whole hog after working really well with one cloud provider rather than multi-cloud. And obviously there's regulatory challenges, procurement challenges and other questions that will come into that point, but that's my point of view. And then lastly maybe a couple of words on skills, capabilities and ways of working. So maybe a minute or two on each of those points. So we worked recently with one of the largest UK retail banks on their cloud strategy, and the CIO there was under enormous pressure to reduce the run costs of infrastructure and operations and had been through three or four successive waves of cost reduction. And he asked us to come up with an assessment of the business case of moving to cloud. And when we looked at the actual, you know, specific applications, specific services that could move into cloud, we said, yeah, there are run cost savings to be had, you know, particularly for more peaky analytical workloads, regulatory reporting, those sorts of things. But actually the business case was quite challenged by the fact there's big upfront migration costs, there's new capabilities to be developed around how you monitor the ongoing usage costs in the cloud that can ramp up quite quickly, and the shift from sort of upfront capex, you know, project-based funding models to the ongoing OPEX based usage models. But when we looked over to the change side of the shop, you know, this is a bank that has a change portfolio of just under £3 billion a year. So a really substantial change portfolio. They had just huge challenges with just basic things like provisioning a server for a project would take six weeks. Every time there's a new project, the project manager has to go out and say, give me a development server, give me a test server, give me a test environment, give me a pre-prod. And every single time it's six-week lead time and sometimes longer depending on, you know, what special variety of technology was needed. And it was pretty compelling that moving to cloud-based provisioning was going to take that six weeks down to, I mean, it could be six minutes, right? But say a week, you know, taking from six weeks down to one week, that's a five-week acceleration for every single project the bank's doing. So if you take your £3 billion portfolio spend, which presumably is delivering more than £3 billion of business benefits, and you say, OK, let's let me take five weeks off of that. I'm bringing, you know, 10% of that £3 billion from next year into this year. I'm accelerating all of my business change. That's £300 million worth of value. You know, that funds an awful lot of cloud acceleration. That funds an awful lot of, you know, open source agile projects. And so we established and convinced the bank that the prize in accelerating business value of change was far, far greater than saving 10, 20, even 30% on some of the infrastructure. And knowing, of course, that those infrastructure run cost savings only come when you've got the last transaction and the last application off and you can decommission it and move things. So that's our first thought. Second conviction, moving up the stack to platform as a service and software as a service. So another one of my banking clients had a reasonably chunky application. I can't remember exactly many 60, 70 servers, something like that, you know, web server application server, database server, typical, you know, multi-tier application, but they decided to move into the cloud as one of their first cloud projects. And they had intended to containerize it and make it a little bit more cloud flexible, but ran into a few obstacles and decided to keep the existing architecture, which required, you know, due to the way it had been built, required a fixed ratio of web servers to application servers to database servers. And they thought, OK, cloud's really flexible. We can move things in. How hard can it be? Well, it turned out that the performance characteristics of the application, once they'd moved it to the cloud, significantly changed the ratio of application servers to web servers to database servers. And because the ratio between them was fixed, they ended up actually having to scale up to 900 servers in order to make this application work in the cloud, which, of course, completely killed the economics. And so I think, you know, this is an area where open source, you know, through Kubernetes and indeed many of the high level platform services that have become very powerful and prevalent in the cloud world is so helpful. You know, just breaking some of those fixed ratios, breaking down some of the monolithic applications makes such a big difference. Another one of my clients, we did some assessment of the root causes behind many of their issues, incidents and outages. And it turned out that more than 50% of their incidents were nothing to do with functional issues in the code. We're simply down to changes in configuration and software versions between what happened on the developer's workstation, what happened in system test, integration test, pre-production test and so on. And there's clearly just fantastic tooling now available in the open source community and indeed from the cloud providers if you're prepared to move from this sort of siege mentality of old ways of working. Which brings me to my third point, which is around the fact that it's not just about the platform, but also about the software development lifecycle. And, you know, very often the move to cloud is associated with an adoption of agile ways of working and it gets very excited about, you know, we're going to go agile, we're going to have two-week sprints, it's going to be amazing. And you have your two-week sprint and then you need to pause for nine months of system test, integration test, pre-prod test, you know, non-functional test. And clearly that's just completely broken. It's completely nuts. You can't work agile if you have to have that kind of siege mentality in order to be able to adopt the cloud agile ways of working. And this is where, you know, I'm preaching to the choir, I'm sure, but it's very obvious that the open source community has developed fantastic capabilities at different levels in the stack from infrastructure as code, terraform, Ansible, Chef, Puppet and so on up into container orchestration where Kubernetes I think is really should be seen as the minimum level in the stack that people should be aiming for. I was really surprised, a bank, even as recently as a year ago, 18 months ago, was saying, you know, is Kubernetes really well-established? Is it well-recognized in the industry? Is it definitely the future? You know, I was astonished that, you know, reasonably mature IT organization was still questioning whether Kubernetes was well-established out there in the marketplace. And I felt you need to get out more. And, you know, it clearly is well-established, not only, you know, as a platform in its own right, but as an interoperable platform, you know, between different cloud providers, which we'll come back to when we talk about multicloud in a minute. And then, you know, monitoring in other levels in the stack. And one of the seminal articles in the world of open source is Eric Raymond's article and then book, The Cathedral and the Bazaar. And I'd like to share my sort of 2021 take on what he wrote, or at least an aspect of what he wrote, which is that in agile and cloud adoption, you want to empower your individual squads, your individual value streams to work as independently as possible and as rapidly as possible. And my assertion would be there's a paradox in that that you have to be ruthlessly prescriptive, actually about a small number of things in order to enable these teams to have the freedom that they need. And a colleague of mine, John Webster, sometimes describes this as the choice between being Catholic and being Buddhist. He says, there's, you need to be the Pope and define top down, thou shalt for a small number of things, you know, things like, OK, what container orchestration tool are we going to use? You know, what source code, source, what, what, how are we going to do our source code management orchestration? You know, what monitoring fabric are we going to use? You know, you don't have to define what editor people are going to use. You know, you'll start a religious war there if you try and go into that world. But if you define a few things very prescriptively, then you can give the freedom to the teams to say, OK, I know that if I change this, I'm not going to break my neighbouring component or my upstream or downstream service because I am, you know, I know that when I check this code in that Jenkins is going to automatically kick off and run a set of automated regression tests so that we can all have confidence that I haven't broken the last known good built. And I think having the confidence and the courage to prescribe a few things in order to enable rapid throughput is so important. Another one of my clients, we did some analysis. We found that the average time from unit test complete on a developer's workstation to beginning integration tests, I haven't done anything, just beginning it, was 86 days. You know, if I asked you to go back three months to, you know, what Excel spreadsheet or code or PowerPoint or whatever, you know, you might have been working on three months ago, you know, probably all of us would say, I have no idea what I was doing. You know, and if you said, you know, somebody has made 100 or 100 different people have made other changes to that spreadsheet in the last three months. And now I'd like you to find out why the formula in cells C 97 is not working. You'd say, well, that's, you know, completely ridiculous. What a broken way of working. And yet, you know, there are lots and lots of organizations that continue to work that way, rather than having, you know, at least a good overnight build, or if not, you know, a regression test every time you check code in. It makes such a difference to the speed at which bugs can be identified and fixed. And so you actually paradoxically get by moving faster, get higher quality code. Whereas normally people think, well, if you move fast, you might break things. Actually, there can be a win-win. One cloud versus multi-cloud. If you're after a provocative, even snarky view on the subject, I don't know how many of you are familiar with Corey Quinn. He runs a AWS pricing consultancy called the Duckbill Group. And he's great on Twitter. And he basically says, you know, multi-cloud is a recipe for setting your whole infrastructure on fire. And I wouldn't go quite as far as that. But I would have some sympathy for his point of view. And I think having worked with a number of banks on their cloud journey, you know, most large complex banks find it takes them, I don't know, nine months at least to get a new, you know, cloud relationship up and running. And often if they're trying to engage with multiple different cloud providers at a time, it can take 18 months, even three years, before real production workloads delivering value. And that's just far too long. I attended a meeting recently with Andy Jassy, the new head of AWS, and he was, you know, loudly proclaiming that GE, one of his clients at the CIO there, had set a goal of having 50 applications moved into cloud in the first 30 days of them adopting AWS. They didn't quite make it. They got to 42. But, you know, he was very proud. That was, you know, a pretty spectacular track record. So it is clearly possible to move very fast into the cloud. But I think the complexities of regulatory approval of, you know, legacy tech complexity and so on in the banking world often, you know, create large obstacles to moving fast. And often that's not so much the tech itself, but it's also things like the ways of working. You know, if you've got a relationship with an outsourced provider that says every time you want to make a change to your infrastructure, you've got to fill in a request for service, you've got to wait 48 hours before they come back and said, oh, that's a special service. You know, you're going to have to have that priced up specially. There's this sort of intrinsic ways of working challenges that also get in the way here. And I think there's, you know, if for each new cloud provider you've got to figure out how we're going to do provisioning, how we're going to do identity and access management, how we're going to do incident management, how is that going to integrate with our on-prem world, you know, it's a lot of work to get it right. And if you're trying to do that with AWS and GCP and Azure and, you know, Oracle and IBM and everybody else, God forbid, you know, then it's an overwhelming amount of complexity. And even though everyone likes the idea of not being tied to a particular provider, not having vendor lock-in, in reality you can find that the complexity increases, the risk, the execution risk increases, and the benefits of multi-cloud can seem rather illusory. And I think, again, open source is playing a big role here in reducing the extent to which vendor lock-in is really a problem. You know, if your containers are Kubernetes compliant, you know, it is actually pretty straightforward to move things, you know, for one cloud provider to the other. You know, if you really, really wanted to, you could do it dynamically, but I don't know if any banks are doing that live in production, you know, it's theoretically possible. But the, you know, the important thing, the thing the regulator would want to know is that, you know, if there were a big problem commercially, technically, however else, you know, how difficult would it be to move things? And, you know, this is where I think at the risk of offending the regulators in the room, this is where I think the regulators sometimes don't help themselves. You know, just if you look through the regulatory guidelines on the use of cloud, a lot of the language is framed around the idea of material outsourcing. And that's not stupid, you know, clearly moving to cloud does have an outsourcing dimension to it, but it's often not considered in terms of the relative risk of the move. You know, if I'm in a picker bank, if I'm Lloyd's bank and I'm moving a terror data application into AWS, right? Lloyd's bank's market cap is what, 30 billion, terror data is 6 billion, right? AWS is 1.6 trillion, right? I've got an existing app that's on terror data. I'm completely vulnerable to them if they go bust. And yet, if I want to move that application into the cloud, suddenly I have to jump through all of these hoops to prove that I've got a way of coping if Amazon were to go bust. I mean, it's just a bizarre way of thinking. And I mean, it's not a completely stupid question. You know, the relationship perhaps with Amazon could go wrong. I do need to have a plan. But, you know, is my service more or less reliable if it's running in AWS than if I'm on, you know, a terror data, you know, old style, you know, proprietary database environment? You know, I would argue that it's actually more resilient, more flexible, you know, say in a Hadoop or equivalent cluster running on AWS. And why should I have to jump through many, many hoops of approvals internally and with a regulator in order to be able to move it to the cloud? Fifthly and finally, cloud skills and capabilities often end up being a key enabler of the move to the cloud. You know, a number of my clients are adopting cloud technology, not just because it's, you know, of the benefits we've talked about, but also because they want to attract and retain the best talent. You know, we heard at the beginning of one of the earlier sessions that, you know, that's a really important challenge for banks. If you want to be able to compete with the web giants, you want to be able to compete with the fintechs, you need to be able to demonstrate that we're working with, you know, cool, exciting technology. And so actually a move to the cloud can be a great way of demonstrating to new young talent that if you come and work for us, we're not a funny daddy old bank working on, you know, ancient technology that won't help you with your career. Actually, we're working with some leading edge technology. But in contrast, you know, most banks today actually have had to skill up for their existing landscape. And so there is a real steep learning curve here. And I think the positive side of things is that because the economics, because the way economics of cloud service provision work, they have this, the cloud service providers have these very attractive recurring revenue streams. They're typically highly motivated to help train and train your staff. And so they will throw or be very generous, shall we say, about access to their training, their certification processes, providing that you're willing to make a commitment of a reasonable amount of your workloads to move into the cloud. So I'll stop there. And those are my five convictions on how banks get value from the cloud. The value is about the speed and cost of change, not so much about the run, that you don't get that much from lifting and shifting your existing tech into infrastructure as a service, but moving up the stack to PAS and SAS and making the most of that great open source platforms and enablers that you have to change your software development lifecycle and your ways of working to move from the siege project mentality to continuous integration, continuous delivery, which by the way doesn't have to be the same everywhere in the bank. The value is typically in the channel systems, in the analytical systems. If you want to have daily releases there that say, OK, our core banking system, our payment system, monthly or quarterly is totally fine, again, that's one of the benefits of the new ways of working. And then getting real workloads, real value delivered on one cloud provider as I would push for way ahead of chasing the multi-cloud, even if it means having to work hard with the regulator and with your internal procurement and other approvals and invest in those cloud skills and capabilities, including all the great open source tools that are available. I hope that resonates. I'd be very happy to take questions, maybe over coffee. I don't want to keep you from your coffee break any further and looking forward to joining you for the rest of the day. Thank you very much.