 Hello, yeah Hello, everybody. So thanks for having me today So my name is was seems wally. I am a cloud solution architect working with me with merantis and Today we will be talking about building a framework for a big and for decisions Before to start so we are having scavenger hunt with the merantis. So just collect these QR codes and Come to collect your prizes at the bill garden, right? I will be displaying it again at the end of the session. So It's no rush so the shopping list so basically we have met with a lot of customers and A lot of them are using so VMs or containers or both of them Of course, we have other solutions, but this is what we see the most So usually we are using VMs when we have like a lift and shift So customers they're having like servers and they just want to optimize them So they move everything to do VMs lift and shift without Doing anything. It's like minimum effort, right? And it works very good It is also used while using like a non-standard HTTP or HTTPS protocols We see that especially in the telco. I have a colleague here like telco expert He will have another session to discuss about it Especially with like seep and diameter and GTP protocol where we really we really need like point two points So this is much easier to use VMs And also it's used in order to run like different operating systems Especially like if we develop with Windows and Linux like with dotnet or these kind of things. Yeah, it's much easier So of course there are other reasons, but this is what we see on the field in the other hand so a lot of our customers are using containers as well and Especially if they start from scratch. So of course if you have a code that exists from 10 years, which is like a Buck free so you don't want to risk it and to create something from scratch But if you are new company, so of course, it's easier in order to build something from scratch and use the All the cloud native and like a depth secops Principle and these kind of things and of course you could put much more containers on the same server So it is also optimizing the servers But really what we see on the field is that people they have a lot of imagination, right? So sometimes they use Pots where they where they should use VMs, especially in the telco field So 5g is moving to Kubernetes. So everything is moving to pots and We have seen that customers are just doing like a lift and shift From VMs to pots. It means they have like huge pots Which are not cloud native, but well it works and We have seen in the other world like people that just stick to VMs So they use VMs like they are pots like Kubernetes pots It means that in case of issues they don't jump in the VM to troubleshoot it They just kill it and start a new one in case they would like to do an upgrade They just start new VMs and they kill the old VMs. They are really using in very smart way VMs like Kubernetes. So at the end, so don't put yourself under stress. So should I move to containers or not? It's yeah, it's like use your creativity lose your imagination and you could achieve a lot of things and It's the infrastructure that should follow you and not the other way around So now that we discussed about the VMs containers or both of them. So actually where to deploy? The first thing and the easiest of course would be to go to cloud So it's very easy with your credit card. For example, you could you could like deploy on cloud It's it's ideal for small deployments For startups and for irregular workloads. For example, I met a customer who told me every morning They start like 100 VMs They collect a lot of data from internet and then did they like make reports for their customers and after two or three Hours they just shut up these VMs. So like the public cloud is perfect for them, right? So however the downside is that at the beginning it's easy and it works well. It could scale. It's like in limitless But it has a higher OPEX and people are talking about like bill shock It means that they give especially if they have like a big enterprise They like give accounts to everyone and after the end of the month, they end up with like millions, right? Also, so someone could ask question about security and sovereignty So they say that data is secure and I believe so but where is it? Is it behind your firewalls? So how many security guards you have in the data center? So all these kind of questions so someone could ask right, especially if we are government or banking or these kind of companies And also there is vendor lock-in. So it's easy, but when we start using it It's kind of hard to move away from it In the other hand we could see like on promises Here, of course So we have very good security and sovereignty. So everything is behind your firewalls You you have exactly like the good networking Then like you separate the back end from the front and from the back end relay So you design it exactly to meet your your like security requirements And of course you have sovereignty, you know exactly where is your server? Where is your database located, right? It has higher capex because you have to buy everything you have to rent the site or buy the site rent the It's like buy the servers You have to buy everything, but if you do the math after five years you would see like the the TCU the total cost of ownership is Lower is much lower and this is like something that we do with Mirantis We do the math the calculation and we see that people are really having back a lot of money And of course, it's yours. So you could like design it to fit Exactly what you need you put the services that you need my it's yours So the downside and I think that this is the biggest downside is that you really have to Right size it and it's kind of hard. So if you have if we have here like some infrastructure I would say responsible so they know about it Usually we go to the project managers and say, okay So do you have like some projects coming in the next one two or three years? And then you convert it in the amount of servers you need and then you buy the servers and if for example, you would need 100 servers and you put 150 it means that you are paying 50 servers maintenance cost for nothing. It's a waste of money and if you put less So this is even worse because you will go with overload So this is I think the hardest point that we see with our customers Also, you would need like to handle everything like physical site cabling cooling energy In my previous company, I could even download AutoCAD in order to know exactly where to put the shelves And also you would need multiple skills. Of course, you need like the guy who's doing the cabling is different from the Cooling is different from the energy. It's it's like a lot of people that are needed in order to To maintain this and promises infrastructure. There is a solution in between It's called like a co-located cloud or metal cloud So this is like something that we see more and more. It means that instead of deploying In cloud like AWS or Azure, whatever you rent servers on cloud like with equinix metal, for example So you have your own servers and then you put on top of them the needed infrastructure like Kubernetes or OpenStack or whatever and sometimes like with Mirantis you have Actually, we have very good partnership It means that you go to this partner and he would take care of deploying everything and you just use your cloud like it's Some public cloud. So this is very good alternative. It's between like the cloud and on-premises So you don't need to wake up very early in order to get the servers which is happening nowadays So they already have thousands of servers And you reduce the cost It's also very easy to use because it's based on APIs. So it's like a self-service in order to wake up a server You don't you don't need to go to the APMI and pixie and these kind of things You just need to go to some API in order to wake up like thousands of servers, right? And so and if you do the math, it's kind of It would lower your cost if you compare it to some hyper scalar But of course, it's a bit more expensive than On-premises network and The good thing here is there is no of under lock-in because you just have metal You just have like servers and networking and these kind of things and then you deploy your own solution on top of it It means there is no of under lock-in and about security so it is It would never be as secure as what you you have behind your firewalls Of course, but is it's much more secure than The normal public cloud because here you could have your own tenants You could have your servers you secure them like you are You could have as well your own like Private networks it means in terms of security and sovereignty. It's better than public cloud If we would like to push things further We could talk about About hybrid cloud So here, of course you gain in terms of agility It means that if you went to see this program manager saying that listen do you have project coming in the next Few months he would say no and then he comes back. He said oh listen. I have actually a project would take Like hundred servers. Yeah here if you have hybrid cloud you could like offload this extra load to the hyper cloud and this was would help you of course to Right-size your on premise cloud. It means that your on premises cloud would be used for the regular workload and if you have some project which is not Which was not planned or some peaks Especially I have seen that in telcos actually all the telco network is Sized for the peaks it's like a size for a Christmas at eight and then All the rest of the year it is like running very low which is a bit of pta So this hybrid cloud would really help you to size it correctly, right? The downside is that you need like your on-premises skills, but you need another team like handling this How to say like Public cloud it means that you need more skills and you need to maintain a different kind of environment and If we push things even further We talk about multi-clouds in this case We are having our in-premise cloud. For example, we are having not only one hyperscaler, but multiple hyperscalers It has a lot of benefits, of course for example avoiding the Vendor lock-in because as you are using multiple Hyperscalers it means that all your Development is is used to work with the different hyperscalers. It means that you do not have Vendor lock-in. It could be used as well for cost arbitrage. So today maybe AWS is Maybe cheaper you could use it tomorrow. Maybe Azure is cheaper. You could use it This is like something that I have seen in telco as well Where they carry voice so every day they switch from one carrier to another one in order to reduce the prices So they have this possibility and of course we have seen that Even with the bigger hyperscalers from time to time. It's low, there are outages So It could happen. So in this case you just switch to another one The downside of this multi-cloud is if you think that hyper-cloud is difficult So multi-cloud is even more difficult because you have to maintain multiple IPIs, multiple like monitoring systems And it's becoming very complex and everyone knows that complexity is the enemy of security, right? So it makes things even worse in terms of security This is so just to give you an idea about Key components that someone might have with the infrastructures So we see that we could have like virtualization Word you could do it for example with OpenStack, right? We could have Conturnization world we could do it with Kubernetes or swarm or there are other solutions But the idea is that we should have one single Orchestrator in order to handle all this word, right and while doing This infrastructure we should really look for LMA like a login monitoring alerting And billing and we will talk about this question in the comment slides more in detail And also so someone should not forget about About the registry repository and the whole CACD that should be secured This is the idea so at the end what's an ideal cloud? So this is like a vision So if there is one slide that you should keep in mind maybe this one So An ideal cloud is like Smaller or a downsized version of of hyperscalar which is on promises It should be agnostic to the underlying layer so today we have OpenStack we have VMware we have I don't know like a Kubernetes we have swarm but tomorrow who knows it means that While building your ideal cloud you should really keep in mind that it should be Agnostic and you should like to all your development in some agnostic way. It's very important Otherwise in five years you would need to rebuild from scratch, right? It should be also designed By having the the time to market in mind very important because this is what would give you I Would say a competitive advantage so against your competitors, right? It's always the the the time to market and in order to achieve that It's very important that it should be like with the unified access It means if I have to do something I should not be log in there log in here It should be like one access and it should be with the with the self service I used to work in some Companies in the past where in order to have your for example database a service You have to go to the networking team to to get some scopes and then to the firewall in team to open the disc ops and then to the I'd say To the DB team and there and this and it would take like a few weeks It would make it project where it should be very very easy. So it should be really like a Self-service Also, it should provide from scratch Most of the services that your company need For example, if I need like DB, I don't know like my sequel or Cassandra or whatever It should be by self-service within a few clicks I get my database should not be like a project and go into the team and it would take like weeks It should be within a few clicks so same for like backup as a service and same for VPN a service and for all the Services that you might need we could see that we could really automate them And it should be used like like directly through self-service Even the security should be automated I would say like 80% of security could go through some templates and CI CDs and like DevSecOps principle And it is it is like something that could waste a lot of time So so we should really look for this kind of automation and There is like something else which is Monitoring so we should have very good idea about what's happening in our platform So is it working well? Is it like at 80%? I should like scale it We should have all the alarms at all like levels like the The networking the servers and what's on top of it? So the full stack should be monitored and Also, there is like something which is very important But it's very hard to do which is having a true granular billing Because today if we go to hyperscalar we know how much it would cost me a project it means that I could see that like This project is costing me How to say like I would say hundred VMs. This has a certain cost But if we do its On-premises is kind of hard because we know how much the infrastructure is costing So how much the side is costing how much the cooling how much the electricity how much the servers But we we cannot know and it's very difficult to know how much really a VM or Some Kubernetes cluster is costing within this specific On-premises infrastructure and this is very important to implement And also it should be hosted So on-premises of course But as we said, it's very important that it should be Hyper cloud in order to write sizes and to offload all these peaks and the special events and in order to How to say Not oversized so in order to implement this Ideal cloud so basically you have two possibilities So either you do it yourself with specific team so you dedicate Specific cloud team in order to do that and we have seen a few customers that are doing it themself In this case, we have seen a few challenges First one of course is is a talent scarcity. It's very difficult to hire Infrastructure talents and even more difficult to maintain them. By the way, we are hiring with me notice Also, what we have seen is that Actually, we do not want to have outages. This is why our customers they take their best elements And they put them in order to handle infrastructure, which is a misuse edge And it's a pity because instead of doing like innovation and very interesting things so they are They are just stuck with infrastructure, which is really a pity and Also, what we have seen is that so as it's not there like day-to-day business. So usually So the software is outdated. It's not maintained. This would lead to outages to to to like Security issues and these kind of things and also like implementing The need that the telemetry is kind of hard so people they just implement something very easy With nodules or these kind of things which is not enough in order to view the full stack So these are the challenges we see so if you would like to do your infrastructure your own So you have to keep in mind these kind of challenges because you might run in these same issues The other alternative would be to go with the partner So like me on this for example in this case you have a few elements that you should keep in mind So the first element is a very rapid and quick deployment It means that from the time that I have all the needed requirements I have like my servers which are Connected in shell so how much time it would take to Have something up and running usually it's like a couple of days, especially if you go with the Reference architecture of your partner. It means that they tested this at their in their labs They have like hundreds of customers using this same like reference architecture So the idea is that it could be very very quick deployment This the second thing that you should be looking for is Would it really reduce the The overhead it means that We are talking about like day two. So this platform is up and running, but it needs to be maintained Upgraded we need to look for the life cycle management. So will this partner look for this life cycle management? so Do they have like automated tools? So how it is working? Would it be easy? So So here it's a I should be like a user of my own platform It's just I go to my platform and I start consume VMs containers and And services The other point is So this partner should help me to bring back my workloads today. I'm using some cloud I am in this vanderlaken So usually these partners they have experience by moving back the workload. So this partner I should so ask the right question to see whether they have the needed scripts whether they did it how is it working and Yeah, this is very important in order to help you to move back all the workload and Also, so today you may be leveraging VMs or containers, but you should look for someone who has both basically Because today you might be having like VMs, but tomorrow Maybe you would need to leverage the cloud native principles. So you would like to move for example to Containers so your partner should offer this possibility to have both at the same time So in summary So you might be leveraging like containers or VMs. You might be having like on-premises or I'd say like on On cloud or both so So just reach out We are in booth B2 and let's discuss about your journey and See how to share ideas because it's very important. I mean the most important thing is the idea and then we see how to implement it so thank you for this session and Let me put again this scavenger hand. So if you would like to Yeah, take it. Thanks a lot So Hello, everyone I'm Elvira Garcia and today I'm going to talk about how to contribute to neutron a one-on-one session I've been working at Red Hat for a year and a half and I'm going to try to get you some Along some of the key things that I think that helped me to get Introducted into the Neutron project So what is this talk about first of all, I'm going to make a brief introduction To the Neutron service and what it's rolled in OpenStack Then I'm going to list you some of the reasons that I think there are To contribute to this project or really to any OpenStack project or any open source project And then I'm going to try to tell you Some of the key things in order to get to fix your first bug in Neutron So first of all Neutron is the OpenStack networking service in OpenStack There are many different Services that allow for many different capabilities in the cloud Some of them are optional and some of them are core Neutron is one of those core components because it allows for all the wiring of the Instances that we want to deploy either VMs or bare metal and not just that it also wires the services between themselves the main components in Neutron are the Neutron server that gets all the apicals to the to the service itself The Neutron DB that has all the information about the networks that we defined because since we have Software defined networks. We need to save the state and the definition of all the routers the networks the subnets that we create and Also the plugins and drivers The plugins are you could say where the magic is it's where the capabilities are really coded and one of the most Core plugins inside of Neutron would be the ML2 plugin that is a framework for creating all kinds of Link layer resources like networks as I said before different kinds of networks and Inside of a plugin you can find Different ways of implementing that capability in the case of the ML2 plugin We find ML2 OBS, which is the open v-switch backend for the ML2 plugin and ML2 OVN too that it works with OVN as the SDN technology on the back So why I think it's a nice thing to contribute to Neutron. It's really straightforward The first reason is because you can it's not with every project that you are able to actually get interested or find a problem and just go into the codebase and Learn a bit about why something is failing or how something is working And if it's because of a problem you can even go ahead and submit a patch for that I think that allows for higher quality of the code and also for higher reliability Because there's more people looking into code than if it was closed source and There's also public peer reviewing this means that for a comic to get into the codebase There must be several people reviewing a patch and at least two core reviewers Stating that whatever you coded into the project is Right for to go inside So the next thing I think it's because you want to learn so you might just be interested into Networking for example in my case when I arrived to Red Hat and I started with OpenStack, I only knew what TCP, UDP and all that kind of really basic stuff and Once I was Here for already a year. I've heard much complex concepts like for example VLAN transparency or trunk port and