 So hi everyone and welcome to our talk The session will be about our journey at orange from big data environment to application delivery, of course using platform. So just before we start short reminder about the fire exit announcement, of course So please notice the different exits that are behind you in this room and In case of emergency just run as fast as possible. No, I'm kidding please currently exit the room and then follow the signs and the safety safety crew. Thank you So a few words about about us because this presentation is done in pair with Alexandre Vasseur from Pivotal So Alexandre is in charge of the platform architecture for the Pivotal customers mainly in Europe. Okay, and among these customers, of course orange. I Won't say anything about Pivotal since there are kind of rock stars in the summit even more since this morning and the and the announcement, but let's have few words about orange So for those who don't know orange actually were a big telecom company initially in France, but more generally in Europe and Middle East and Africa also and we are dealing with very varied activities that start from selling cell phones dealing with networks And up to many other activities like dealing with software of course and that's also why we have We have put a lot of efforts on the credit fund re Since we are also a silver member starting last year so Since we have very varied activities and very different business units actually we don't have one CF platform at a range we have several of them Some platforms are actually open source. We have Skeet Center CF Skeet Center at orange that is handling all this part and we also have some platforms that are Pivotal Cloud Foundry and Today actually we will prevent the Olaf which stands for orange lean application facilities that is based on Pivotal even even if we are currently Let's say broadening it with an open source platform in parallel. Okay, so Let's go to business And let's go back 18 months ago Actually in our team we were mainly focused on big data activities So it means that we have our strengths on big data. We are the very big adobe cluster Lot of nodes lot of use cases and so on But we also had some weaknesses The fact that for example We did not have much applications in productions simply because it was a pain to put applications You know to host it to deploy it so We had this kind of weakness, but we knew also our ambition the ambition was let's say to do as most of the Big actors are on the internet and be able to deploy very fast and very frequently in the different software simply to be able to To develop it based on the customer feedbacks So that was our ambition and since we have the big data we we thought about enough for that is kind of improvement loop Which is based on the four steps the first step is first to get all the Data from the customers and put it in the in the adobe cluster Then we'll learn some analytics on top of this and get the customer feedback customer trends and so on on top of this we will Build let's say some hypothesis and develop new features or even new new software The last step of this improvement loop is of course to deploy the software in production And again on this part. We were not very well Very well equipped. So that's why we launched the Olaf project. So So we had kind of our main requirement which was since we are IT engineers and Most of the time it engineers are lazy people Our main requirement was we just want to push our code into production But since we knew a little bit about the cloud native applications. We knew also that it will be possible for greenfield applications or at least applications that are Cloud native of course So that was the first requirement The second one was okay. It's good to have this kind of platform. It's even bitter better to have it Inside the orange ecosystem I'm meaning here, of course the big data All the Git lab pipelines for example or SMTP relays and so on so Actually, we have a lot of legacy and a lot of tools in our ecosystem So that was mandatory to have it in our private infrastructure The third requirement was more on the upside because I I've told a lot about the development pushing application But once it once it's deployed it's even better to have the application monitor and have tools for apps and up and platform operators So all these requirements put together actually we found that the cloud foundry was the best solution to have all the answers to these requirements I've put some of course some items that made us Make our decision, but I guess most of you know about the different functions of Cloud Foundry So next step so we we chose Cloud Foundry that was our tool then we built and designed the High-level architecture. So here it is a very high level But all the Olaf components are in blue on the left of this of this shema So first the first step actually was to build our cloud provider and here We decided to put it as a private To make it as a private cloud provider based on the pun stack. Why did we do that for several reasons the first one is that Data is very sensitive. So since We were using some data. We didn't want to put it on public cloud Okay, and we had to to have it as as close as to possible To the I do cluster. So it was the first reason of the private cloud Why we chose open stack simply because first it's Kind of recommendation at orange or at least in our in our department and also because we have a lot of People and a lot of knowledge and type of open stack. So let's reuse what what we know and That was I guess the right the right choice Second second component that makes Olaf is of course the Cloud Foundry. So here are the POTL Cloud Foundry and Then we have This big big components and which is always evolving which is the marketplace services So here you have of course some database services messaging spring and so on but we have also some specific services that we built Like the like the big data connectors. So currently we have I've and HDFS You can see that among all these services Some of them are from POTL other services are from the community So we simply get them in and install them on the on the platform. So just some integration task for us and The last type of services are from from orange. So directly developed by a by our teams Next component from this architecture is the VM on demand Actually, since we had our own open stack, we also decided to to give functions to our to our developers and saying, okay, if you want to Push for example your application that is not clone native Or if you want to use some services that are not in the service marketplace in this case Will give you a VM a bare VM and just use it like this Of course if we have several customers that come and ask the same service the same service Then we'll add it in the in the marketplace. I mean, this is our strategy Last component is a non-boarding web portal. I will give you some details later. So let's Let's move to to the next on the right, of course the big data and The orange ecosystem that was already that were already existing before we started this project with a lot of integration, of course so What about our journey with with Cloud Foundry as I said as I said once we have the we have Chosen the the Cloud Foundry tool or the Cloud Foundry solution We made all the design of course and then there was the installation and implementation of the platform For this we relied a lot on on Pivotal because again, it was very fresh for us and And we use the dojo from Pivotal to do that Once done we focused during several weeks and months on the core team for this for this project to maintain it so in its It consists on both people from architects, I would say IT engineers either from Pivotal or from range Developers and of course some ops Very small but very strong team for for ops Next step was of course to enrich the platform with some some bush releases from the community So there will be a slider slide on it And also for the automation so all these steps will be detailed later And last but not least because yesterday we we heard a lot about Product platform as a product and so on or to onboard the customers actually we have done a lot of things on this So it will be the end of the presentation So Alex Yeah, thank you Olivia for for this journey. It's been interesting So when we started working with that team and Olaf project We really had something in mind and wanted to have them You know behave as the you know platform as a product team, which is like kind of easier than said, you know than done Especially, you know considering the very large company very large process very heavy design decisions made up front Obviously, you know at Pivotal, we are very opinionated. We're saying installing PCF is easy We have a bunch of tools and practices around that so working with customer to kind of Of you know focused on the end state which is never ending actually It's a kind of a loop in having that platform having a road map and having upgrades and having better capabilities Measuring your own sort of service level indicators and improving the marketplace kind of was built into our Kind of DNA and and and how we wanted, you know orange to behave So the way we've been working is by deploying a bunch of pivotal people working like full-time Several weeks with the orange team and building that kind of core team having us a backlog and Kind of design decision made not not with you know too much thinking but you know very quickly Doing things right like fail fast So this is like doing a spike if you have a doubt about something just try it and then just like slide away Right rather than having kind of too much discussions or you know too much consensus going on in discussions This said you know at the scale of orange. This is kind of easier said than done as I said they have you know Four location two countries different culture. They have a lot of you know Teams around that like the security team the network design team You have you know architecture validation committee in that process as well So it's not always easy to do that But I was really important to you know start to have that that culture And obviously they had no platform before and after they would have a platform for for a little while And so they had to organize their self, you know as a team to kind of you know take that platform forward So that was important one of the example was how we used the open source Community components to enrich the platform and I'll give you a few examples there We started at some point like a like a whiteboard on how we should monitor the platform You know, what's the cloud ops team would need to Have metrics and dashboards around around the platform and operate PCF So we started with the whiteboard we ended up with a like half a day whiteboard seven different solutions and no decision Right. So what do you do from there? So we decided to recommend orange to use the you know, Prometheus Bosch Elise Which was initiated back back then by Stark and Wayne We had experience with it within the people team had pipeline to actually deploy it Into the into the platform as Bosch Elise And so from the ID to the kind of Prometheus in production on their platform was just two weeks And then we had feedback from the ops team about you know how it was You know pushed as a as a Bosch Elise and and school pipeline and and what's the value they got from the metrics and dashboards And it's been it's been kind of one of the tool that they decided to keep rather than you know go back to the whiteboard Did the same leveraging the log search kind of log as a service providing log as a service to the to the application team and also did the same for Providing disk as a service even relying on open stack. You could have you know S3 storage provided to the application Using service workers, etc. But while they were building that capacity in open stack We also enabled the Bosch Elise for NFS server and having disk as a service Attached to the garden container, you know, which might you know help for you know Application that do need like kind of disk as a service inside inside the container So there was a great way to kind of unreach the platform with additional value and have feedback loop Based on that work There's there's a lot of opinions built in when we work with people with cloud foundry. Obviously, you know You know reference architecture, you know at pivotal we spend a lot of time on reference architecture on You know open stack or other You know cloud cloud cloud providers that we we we have So as you know important for having the first design then we Have this ops manager tool. Maybe some of you have been into some of the other a session today To see that which helps the installation, but there's an API on top of that and so what we provide also is pipelines so that as a As a team that is managing the pivotal cloud for me You have pipelines to actually upgrade the pivotal cloud for me very easily and the pipelines are already provided What the pipeline pipeline do is that they go and kind of download Bosch releases or or specific packaging from the pivotal download website called pivotal network And then they kind of propagate the Bosch release down to the system using the API and From there Obviously, we've been working also On making sure the whole practice around the platform was sustainable, right? Because the big the big thing is that we don't stay forever with the customer We're done after a few weeks So they have to run the platform on their own and we're not there I mean we have support team and engineering etc. But we're not there anymore. So they have to be Building that sustainable practice in the ops team and in a way they have to trust Bosch just as we do right like Destroying VMs having upgrades and canaries with Bosch in and making sure you have no downtime Initially, this is like weird, you know, you don't really believe it's true So, you know getting this done and having that as a practice Was interesting in the early days. They were like doing manual upgrades lots of planning before the upgrades and You know after a while they decided to have like full automation especially for patch upgrades In addition having this like pivotal cloud for me and open source ecosystem Marketplace was pretty interesting. They orange has submitted some of the work into the github open source that they have And you're having the whole skill center contributing back to the ecosystem on that and and the other thing that we really wanted to do Properly was integrating to their developer ecosystem right not not providing, you know, sustainability sustainability for the ops team But also for the dev team and the best way to do that is to integrate to the to the pipeline to the kind of ecosystem of the built and release Cycle for the dev team. So we had done integration with your githlab on that aspect and having kind of end-to-end automations with the application development team release cycle So they're essentially using like concourse for the platform automation and githlab that they were already having at the global scale for the Development deployment aspect of it. I think all of you you wanted to talk about How to communicate to the development team the value of the platform? Yes. Thank you Alex Actually, we discussed a lot of technical topics up to now, but as I said, there are many things about How to go on board customers and how to make it really a platform as a product? I mean actually it was it is a product for orange orange developers again We started from almost nothing from Unclut for me again 18 months ago. We didn't know about about CF except that it was a buzzword. That's all But now what we want to do is to to help Most of the developers to develop this way So it means currently we have hundreds of for users and we tend to have thousands of users Okay, we already have some great results Dorman yesterday she was Saying okay, it's important to get some KPIs. Yes, and we already have some KPIs about some Some projects internally that are saying, okay, we have much increased our velocity and so on so it's I mean It's nice, but to do that. We had two options First simply give them the tools. Okay, the apps manager CF CLI and so on and say, okay Just Google your questions and you also you will have some some answers or the other solution that we took I mean that that we get that we took was to say we need to onboard these developers because Actually, they don't know about about cut foundry. They don't know about the the cloud native strategy and so on So, how did we do that? First by providing some support. Okay, so it can be technical support about how you will use Clubs foundry. Okay It is also help about how to design an application in a cloud native way, of course We do a lot of promotion and seminars and demos and so on ternary It's very important for further. Let's say to make this this platform known as much as possible for orange and last thing is the onboarding Web portal that is here Actually, it could it could look like a very simple web portal and actually it is but the information inside is Very important for our developers. So the first category here is all the offer description So explaining what is an application runtime because yeah, we are using for the moment only the application runtime What kind of services they can use? All these service marketplace is very interesting the build pack and so on Then we have put a lot of work on the tutorials again here. We want to put in a single place Every tutorial that come from how to install the CFCLI or make the first push to Deploy applications using the different micro services on spring container to container and so on so yeah Tutorials again are evolving almost every day We have also decided to make the user Creation and authentication on this web portal. Why did we did we do that first big? I mean essentially because we wanted to provide to every developer a single place to try Clutronry means that when the user will create its account Actually will create a dedicated space with limited limited resources, of course But then it can just try it as a sandbox for us it was also very important to give it easily to all developers and Then once the user he created we can simply It's a transfer it to the apps manager and other tools Yeah, that's that's the deal so About the customer onboarding we have also worked a lot on something that was very very new and very special for us Which was video to explain to our big data users Oh, we will change their way of their way of working. So since we have a time left I propose we can we can display it It's in English. So yeah Should be a large company like orange develops thousands of applications With a great deal of expectations rapid design reliability and adaptability But behind each application There is a huge amount of data a huge amount of coding so implementation can be long and complex And behind each screen there is a Single developer so how can the work be coordinated more efficiently? With the deodd offer Olaf orange lean application facilities a New solution which enables each team to create and manage applications as simply as possible Imagine that developers are given everything they need to implement and operate these applications Imagine that the implementation no longer takes weeks But seconds Beyond the coding developers must also create structure and connect a whole environment with Olaf the environment comes to them All they have to do is to set the parameters according to their needs Take advantage of all its benefits perfect integration with the richness of the orange ecosystem Access to orange groups big data which will help to deliver the killer applications of tomorrow a Stable and secure platform allowing continuous improvement of the applications in response to clients feedback and easy to use an Interactive way of working guarded from day one by a team of experts Choose an innovative work environment which adds a new dimension to the deployment of applications at orange So developers can just develop Olaf by deodd Okay, so again don't see it as Advertisement, I mean it's totally internal offer but here the goal was to promote the platform to our users So it was very kind of weird Things to do simply because we are technical guys So when we launch this this stuff, we didn't see if it will work and actually if we I can give you a Let's say a recommendation If we work on this kind of platform Please promote it to your to your users. Actually, it's it's very important also and That's all for for the presentation. So thanks and If you have any question just one one thing one condition, please speak a little bit slowly for me, of course So it's up to you. Yes Yes Yeah Yeah, it's okay Okay, so actually we have persistent storage using currently NFS and we are switching to to S3 storage for this Sorry Yeah Actually, I think in your architecture you have like the data backplane is separate, right? They have like a separate Hadoop cluster that sits a side of PCF Yeah, right and they have S3 to kind of facilitate data transfer in between and right So, yeah, the kind of PCF still is like kind of a stateless container platform was a catalog, right? Yeah Maybe maybe I can show the yeah Actually because I knew because of the title. I knew we will have some questions about the But the big data of course So here, yeah, the Hadoop cluster is totally Separated from from the PCF platform It seems on its own private network DMZ and so on. Okay, actually we have To make access from the application to the big data we have two ways of doing it one We call native access, which is a little bit complicated and the other one uses using service brokers, okay? So for the native access actually we are using Kerberos for the authentication Give him a key tab and then we provide this key tab To the application to access the big data. Okay, but this way actually Yeah, it's a little bit tricky So we have built some tutorials, but we still need to help the developers to use it So we have also We've developed some service brokers that are using nox this time They're kind of simple simple service workers. Okay, their main action is simply to get the URL I mean give the URL to the to the application so that it can access using HTTPS to AI and HDFS Okay, all right If there are any other question Yes Okay, actually we don't provide this platform to developers outside orange Okay, so if we make a software for a third party it will be developed by our teams that already use the orange ecosystem, but the applications themselves can be either Displayed I mean provided on our intranet or on the internet. It depends I think if I understand the question The question is when you're working with third-party provider that are you know doing staff augmentation for the development team Would they have access to the platform? I guess I guess the answer is yes If they are within orange Yes kind of working in an orange team and joining forces with an orange team Yes, but they wouldn't you wouldn't like provide that platform To an to an external completely external kind of self party development team, right? No. No, we provided to internal teams, of course And yeah, if they want actually we already have some some teams Providing new services, they will provide the service provide the service broker So we can integrate it in the marketplace, but it's only for again for internal purpose We have so many developers and so many projects inside orange that yeah, this is our target Yeah, right. So I think we We have reached out of time so if you don't have any other question Thank you for attending