Okay, welcome again to my talk on Nebulous. I'm at operating system for broken hyperdistributed applications on cloud computing continuums. The talk that I wish to deliver to you today is titled have your head in the clouds while keeping your feet firmly on the ground. And I'm Rodosov Pirisic from seven bullscom and this is my email address at the bottom. The project that I'm about to describe is funded by the European Union and the title. While the only connection between the title and the project is the word cloud, you will see from the presentation that this title is actually pretty relevant to the concepts that I will be showing you as both having your head in the clouds and keeping your feet firmly on the ground makes sense in the various contexts that will be discussed quite soon. Okay, so without further ado, let me show you the agenda. So this is the agenda of today's presentation. First we'll discuss the concepts, the concept of the cloud continuum. Why do we need cloud continuum? How does it relate to other concepts that are found, commonly found in discussions like that? Then we'll discuss what a meta operating system is. I will introducing the nebulous project and give you various details on this project. We'll also have a sneak peek into the nebulous architecture that we are planning. And finally we will have a summary of the contents of this talk. So, regarding the cloud continuum per se, this is very important. This is actually fundamental to this project. This actually has been defined now. So the cloud continuum, this term has been used in various ways, obviously for the purpose of a common ground. I'm introducing it now in this form following the paper with Moroskini from 2022 actually Morriskini atal. It was published in IEEE access and the authors provide the following definition of cloud continuum that this is an extension of the traditional cloud towards multiple entities, and the examples given on Edgefork IoT and this dimensions provides analyzes, processing, storage and data generation capabilities. So basically, if we actually translate it, transform it into other form, the continuum of something is an extension of something, right? And these case this is a cloud continuum. It's an extension of the cloud and towards some other entities that are relevant in this context. So like in this context of having computation somewhere else, these is Edgefalk Iot, and we'll also see it more detailed on some other slide. What I wanted to underline actually did on the slide is the data, and this is actually very important. That data is what is describing the computation. Actually all the kinds of efforts that are being made in that it's actually no computation without the data. All right, so we'll see that in a moment as well. But first of all, why cloud? Continue? Why do we even study such a thing? Why is cloud not enough? So if you are practicing cloud, you probably already know that cloud is not a solution that will cater to all and every need that your application, that your infrastructure, that your solution service might have, right? First of all, we know that these applications themselves, these know no boundaries. Like they don't actually care unless they were written in such a way that they have to run in the cloud. Like for example, we see this lift and shift approached, that people are just moving applications naively, let's say, into the cloud, and probably also taking the cost of doing that this way. But nonetheless, applications, no boundaries, they don't care. These prefer that they have scalability. It's good for the operations to have the using capabilities. So cloud is actually good for that. So considering the continuum from the perspective of having some other resources and joining those to the cloud, you enable yourself to utilize the semifinals, let's say almost finite, less resources of the clouds, public clouds. So usually when I just say cloud, I will mean the public cloud, as we see that offers us the infinite amounts of processing power. So the cloud might not cater to all the requirements that apps may have. So you might have some technical constraints and non technical constraints that are either hard to achieve with a cloud solution. With a cloud only solution, for example, remaining the technical constraints, your latency and bandwidth requirements might be such that the various ways that you need to transfer your data, it might not be sufficient to have all your computing put into the cloud. You might also be using some specialized hardware, I don't know, maybe FPGA. You are doing some scary new algorithms that are optimized for some FPGA, and you actually need to utilize that software to stay competitive with other competitors. And similarly, you have the IoT. Generally you cannot directly connect the IoT to the cloud. You do it via some other means, via some other intermediate provider. Of course you can connect it directly to the Internet and then reach out almost directly to the cloud, but this is usually not the case because of the various ways that you would like to process your data. And this is actually very relevant to the latency and bandwidth that I discussed a moment ago. Similarly, on the side of the non technical constraints, when you are considering compliance with the various policies like the european GDPR, the California CCPA, and the other global ones like PCI, DSS and HIPAA, that is, I think, mostly United States. And you might also have some internal business compliance. And in cases like this, cloud providers usually provide some way to help you with ensuring the compliance with the policies that they are at least aware of. But in some cases, it might actually be the case that you need to keep your data off the cloud, like it shouldn't be conducting the cloud in the first these, at least not in some anonymized, cleaned up form. Right? And in these cases, the cloud only solution might not be the best one. And considering the actual current landscape of how companies are behaving towards the clouds, we know there are companies that have still not migrated to the cloud, that they run off their data centers, they use collocation services, or otherwise avoid the cloud at the moment still. But some of them are looking for a solution that will help them to utilize the cloud and do it properly, not just going with the lift and shift approach, which most companies have already learned the hard way that this is going to introducing a lot of cost for not that much gain out of the cloud. And they would also like to keep their current investments. So they already put some money in the existing infrastructure, and it doesn't make sense to say that, well, this is all obsolete now, let's throw it away. Well, using backs is no good for a company. And some other companies that, like I mentioned already, that have learned these hard way that the cloud is costly, they are actually looking for an optimized way to utilize the resources in the clouds, as well as looking for some insights, whether some of the workflows that they have would better be made running on some other kind of infrastructure, rather than using the clouds, the public clouds. And the underlining aspect to all of that is we could answer the y cloud continuum, because data, now we hear the terms the big data, the Internet of Things, the IoT, the industrial Internet of things, et cetera, that produce a lot of data. And considering these computation in general, when you think about that, even if you consider a game, it's always processing some data. So in case of a game, of course, this is like processing the interactions, the human interactions, against some world state. And basically, this is your data. So the world state and the interactions that happen to it, consider a constant streams of changes to that world, being it a multiplayer game or a single player game. Of course, for a single player game, it would be on a local machine only because it doesn't make sense in general to centralize it. But also in all other cases, you can actually define what data is like. For example, if the user has a form, then the user is inputting some data into the application, into the services, and you can also get the data from other sources, like the devices, the sensors. So you either have a human or you have a non human source of your data, which creates some data provenance, of course. And further, this data that you gather, that you generate, that you obtain, needs to be stored somewhere. It's either stored transiently while it's being worked on. So this is in the working memory usually, and it can also be stored persistently. So this is most often these storage that we think about, that persistent storage that doesn't get cleaned up after some thing fail or distributed, et cetera. Then we have the state of the data that had been transferred. So it's being in transit between two computing nodes that they need to communicate this data to carry out comparison. Or maybe the other node is going to actually store the data in some persistent way. And finally, of course, the thing that takes the most of the computing power is the actual comparison. So we know that of course the various chips need to be working for the generation, for the storage, for the transfer, of course. But the actual processing is the thing that is consuming the computing power. And similarly to the generation and obtainment of data. In this case of processing and analyzing data, it can be either human driven or automation driven. So either it's human doing some computations like running the comments, various comments and describing some results in excel, for example, in excel like software. So the spreadsheets or automation driven, like it's a workload, it's a service that has been programmed to do something this way. And this is the computation. And by adapting cloud continuum, what we are allowed to do is we can optimize these flow, the data flow, the way that we act upon the data. And two these we can achieve the best utilization of available resources. Like the slide says, in that if we disregard the fact that we might have some better environments, some better place to carry out these operations, we are actually disregarding some optimal solution that is impossible when using just the cloud. Like I mentioned already with the IoT case, it might make sense for various reasons to process the data closer to the IoT device, rather than trying to stream all the data directly to the cloud. And where might this continue live? So like mentioned before, we have these public clouds. And these is fundamental for the cloud continuum, because in the name you have the cloud. And basically this is usually consider a public cloud. Of course there are private clouds, but they're comparison to data centers like these really differs in the way that you manage your resources. But this doesn't usually meet the definition of the cloud, as it's for the public cloud, with the resources being almost finite, less, and so on and so forth. In these case, this is often called the fog. So the layer between the edge, the IoT devices and the public cloud is the fog. So whatever we put there, our data centers either managed in a more traditional way or a private cloud way, like with Openstack or some other cloud solution from another provider. And then we have, apart from the cloud, from the fog, we have the edge. And in case of edge also, there are multiple ways to look at it. Either you consider the edge sites, which look like these data centers, but they are actually smaller usually than the fog like data centers. So this is also often called the micro data centers. We have also a similar concept with the multi access edge computing, which is what these telco providers provide, offer to the users, in that it's similar to the public cloud, but running on the telco devices. So usually the public cloud providers don't run the telco, so they don't operate at the level of the end user network, but in the case of telcos, they can do that. They actually do that. And this offering of multi access edge computing is also what is happening now. And similarly closer with each step, it's closer to the potential resources of your data, with the final destination, with the actual source of the current big data, let's say, is the IoT devices. So the devices that are usually not equipped with so much computing power, maybe they have some computing power, but generally not as much as a typical server that is being deployed in data center, but nonetheless, they are there and they're important sources of the data. And regarding the other concepts I mentioned in the agenda that are often discussed together with cloud continuum, and that they have been mentioned much, many more times than cloud continuum so far, because they are older terms, they are also often mentioned now by these public cloud providers, and as they offer the solutions to the multicloud and hybrid cloud approaches to cloud. So in case of multicloud, what they usually mean is that you will have two or more public clouds, right? So pick two major cloud providers and they will probably have a service to make it easier for you to use their services to connect their networks, et cetera, et cetera. Regarding the hybrid cloud, it's similar, but it says that you would have this public cloud and not a public cloud. So this is what makes it hybrid, because it's not really one type of the cloud. Like in the case of multicloud. So we have all of them public. In this case you have one public, the other one is not public. And like I mentioned previously, it could be the case that the other hybrid part is not actually a typical cloud. It's not even managed like a cloud. It's just your data center. It can act in a similar law, it usually falls in this. And finally, this is like a generalization of this concept. So you have the cloud continuum when you just go mix and match as you wish, as it suits your solution, as it suits your product that you're trying to deliver. All right, so finishing with the cloud continuum now, a little slide of the meta operating system. So since you are here, you probably know the operating system definition, at least at these very, let's say shallow even level, but generally know what an operating system is. And these Wikipedia gives these definition that matches it very well what the operating system does. It manages the computer hardware, software resources and provides common services for computer programs. And if you go and try to recreate the meta operating system by studying what the meta means, in this case the meta from the greek meta, so meaning after or beyond is a prefix remaining more comprehensive or transcending. And these, if you build these definition of a meta operating system, which is at the moment not that well defined in general in the literature, you get this definition of a meta operating system. This is system software also, but this manages sets of computer resources. So you have more than one computer, more than one source of computer resources, let's say. And still this system software provides common services for computer programs running on them. So you pluralize this, you layer it on top of the existing operating systems and this gives it the meta flavor of an operating system. And this is basically the goal with the project that I'm about to describe. And here it is. Here is the intro to the nebulous, nebulous as it has been shown on the very first slide on the title slide is actually a name for a meta operating system for brokering hyperdistributed applications on cloud computing continuums. So we have discussed the meta operating system, we have discussed the cloud computing continuums and what we haven't discussed is brokering hyperdistributed applications. But of course it's been all the time in the background when we discuss that we have the data. The data is being operated on by the services, by the applications, in other words by the programs that are running there. They're distributed because they can run on more than one computer. These is where it comes so we have actually already mentioned that and they're hyper distributed just means that well this is going the hyperscale, right? So like not the two computers in one room, but going over distances over many components in that and brokering. So suggesting the optimal solution, finding the proper way, the optimal way to actually utilize those various resources. So we have also mentioned that and nebulous is a consortium project, so it's funded by the European Union and its consortium. Actually consider of 16 organizations, we have both academia and business, we have both tool providers and consumers which provide these use cases. The schedule is as follows. We have three years for these project and it's quite new. The link is nebulouscloud EU. I invite you to visit this site for more details, but I'll try to provide some more for you already. So these are the logos of the various partners in the consortium. You might recognize some of them, you might not, but for the details on them please visit these site. And this is one of the slides that tries to describe how nebulous is being imagined. So what we already discussed is cloud continuum. So you have resources that are in various places, you have the public clouds, your private clouds or otherwise your major data centers, let's say you have your edge devices that are those micro data centers, at least you consider them to be on the edge of your solution. So closer to the actual sources of the data you have at the bottom, your IoT devices, so they can also count as edge devices, but they don't have that much computing power. And what you want to do is that for your particular solution, you want to have an ad hoc infrastructure solution, right? So for your software solution, for your product solution, you want an ad hoc infrastructure solution. Your ad hoc cloud computing continuum designed specifically brokerage specifically for your use case for your application. So mix and match whatever you want. And also, as I mentioned previously, that telcom providers with their multiple access edge, these own private data centers that are very close to the end user connections are also to be selected by such a project, right? Okay, so regarding the various details on the nebulous project, regarding development at resources, the nebulous will be developed in open and it will be offered as open source software and the license will be Mozilla public licence version two. So this is a very liable license, you might know it already. It's similar to the approached license, but it differs in a few bits. So for those interested, you might look it up on the Wikipedia. The development platform is yet to be discussed again because, well, this is, sorry. So it's yet to be discussed because you are previewing this from the recording. I will probably know by the time you are watching this. So you can reach me on the discord or via my mail or otherwise via the project website. But at the time as I am recording this, this is not yet fully established, so bear with me for that. But also I would like to mention that nebulousness is not only about the software that we are producing. So there will be software deliverable, these will be something to plug and play, to configure, run your software, et cetera. But we are also producing other resources, other deliverables, as we say, what we do. We have the open access research, we have the cloud continuum culture, dissemination, this being a part of it. We have training that we offer, and we also have a relevant open call that I will describe in a moment. And regarding the directions of work that we take in nebulous, these are the directions of work. These were also included in the description of the talk that I'm giving now. So first of all, this way to model the actual applications that consider the full cloud continuum, cloud computing continuum, including the folk and edge aspects as I described these before. This will be research and solution to do that properly will be described. Again. This will consider both the infrastructure that you can offer, that you have, that you offer to your application by means that you have it, that those that you would like to use from the public cloud providers and telco providers, but also apart from the infrastructure, you give also the application requirements and the details on the data streams that you will be using with your application. So the data will be driving that. And similarly you define this. We give you a way to define that so that the platform can assure the quality of service that you desire and also optimize it. So the next bullet point regarding the optimization, so that it actually selects the resources that work best for your particular use case. Third bullet point considers that it's important to not forget the security aspects of doing that. So whatever you encode in the application specification should be transferred, should be materialized in this cloud computing environment, in this cloud continuum environment, in such a way that it still considers the security aspects, that the final solution is actually secure by all means. And final bullet point this is something novel that we are also trying out. This is not per se, let's say, specific to cloud continuums, but generally for the services that are being offered is the monitoring and conducting those smart contracts based service level agreements. So the service level is not only realized by its textual form that is being analyzed interpreted by humans, but it's also automatically monitored, recorded, and its violations are detected in a smart contract way. Right. So this is also some relevant aspect that we are pursuing in this project. And the open call that I mentioned earlier is that we have this amount of euro in the fund for nine grants and this open call will open in 2024, actually. And what the open call gives you if you decide to join it and if you actually win this call is that you will be given the opportunity to discover these software, to roll out your software, your solution on the nebulous platform that we'll be delivering, and you will help us. So you will join this in the middle of the project and by joining it you will also help us build a better product by the end of the nebulous project. Right. This is one of the newest additions, let's say, to the european project and finally going something a little more into the architecture, into the technical side of the nebulous platform regarding the architecture assumptions. So what the nebulous tries to do is obviously tries to support the scope of the project. This is like paramount that we support what the scope of the project is in a very coarse grained way at the very high level. Let's say I have already given you a description of the scope of the project by all the slides, by all the previous slides, and this is generated direction so that it's going to cope the various ways to cater for those requirements in the scope. Then the other assumption is that we support the use cases of nebulous. I'm not introducing those because I don't realms have time in these relatively short talk, let's say. Well, maybe not relatively short, but not long enough to really discuss every little detail that is relevant to the project that I'm describing. So for details on the use cases, and there are various, please visit the website that I mentioned previously. And the other bullet points speak about the technical, actual selection of the components that would be used in these open source software on nebulous. So we are reusing the open resources software as much as possible that we have already been contributing to in various ways or using in various ways. So first of all, these open resources components of multi cloud platforms that exist already from the other european projects, for example, these melodic functionizer who have already been completed as well as the morphemic that is nearing its completion now, but it's still work in progress and on the side of orchestration like the consider whatever orchestration that you can do with Kubernetes. And I said on the slide and more, because with Kubernetes you can actually do more than just orchestrate. And the last bullet point is actually all about that. What we want to offer to the end users is that they have this user friendly, declarative, reproducible and GitHub friendly as well, app modeling in Yaml so that it's not scary, it's something that the entire industry is adopting at the moment, right? So having it GitHub's way, having it reproducible, having declarative, basically making manageable realms manageable for the end user. And our approach is very much inspired and based on Kubernetes custom resource definitions with custom controllers. So we have been inspired by the existing crds and we are basing our solution on crds and the custom controllers that we will provide, that we will develop in these very project and these sneak peek into the nebulous architecture. So like I mentioned, this is still early and we don't have many details to show you. Sorry for that, but I'm really open to discussing that with you via mail or via discord while I'm there by the way, I will try to be there, but my time zone might actually prevent me to be active at all times during these actual event. But feel welcome to leave me a message or just leave me a mail and I can certainly talk about it a bit more. And regarding the sneak peek into the architecture in the center, we have this cope that is Kubernetes driven that provides the various services for the actual implement, for the actual realization of an application on this meta operating system. At the top we have the users who provide, as I mentioned already, the available infrastructure description, the application description, the data streams description, and the goal of the meta operating system is to realize the application in the MOS optimal way using the resources that are available and considering these various constraints that have been given in the various descriptions from the user. All right, so without really going into the details of the various points, you can see that on all the sites of the site of the public cloud, the private cloud, otherwise some data center and also the edge fog nodes, wherever these are, we consider that there will be another local Kubernetes cluster that will be somehow federated with the main cluster and via this federation as we call it now, we will opensource that the workloads that are meant to be running in the target environment actually running there. So it's like a two layer orchestration for the nebulous meta mos where the core layer decides on the first level of the orchestration and then the second layer decides on the others. Furthermore, we are discussing considering various elements of fault tolerance, like if one of the sites is actually disconnected from the core, how should it be final that the site should still keep on using and optimized locally for the resources that are available for the site, for the resources that can still run to some extent, like degraded more or otherwise for this local site, whatever that use case is. All right, so things like that is what is the scope of this particular project? First step is of course, given definition of that, what are the exact details, how we want to approach that? And this is still in the talks, so that's what we are doing very much at the moment, that we are actually spending lots of time on that. Well, that's basically it. What I had to tell you about the nebulous project and its state at the moment. And this is what I have a summary for you from this presentation. So I won't be reading it for you because you can pause the video, it's pre recorded if you want the summary. Actually, the summary is bulletization of all the most important points from the entire presentation that I would like you to somehow remember and keep from this particular talk. And with that said, thank you very much for having me and for hearing me out to the end. Radosafijik seven bulls.com ciao.