 Hi, I am Prashant Mishra, Founder and CEO of Click2Cloud. Today I am going to talk about the future of 5G, AI and Edge Computing. The project that I am going to discuss today is Centaurus Project, which we have an open source community, well established community in combination with Futureway, Click2Cloud and several other industry leaders. Today I am going to talk about the design and ideas and the contribution of various community members, how these products can be utilized into the market and the future of these things and your contribution towards the community. Let's talk about what is this project all about. As we talked about the future of cloud native, Edge and AI, in the cloud native space, Centaurus Project is all about hyperscale computing. Hyperscale computing is something that has been originated from the basic infrastructure of Kubernetes. We are expanding the Kubernetes infrastructure beyond the normal limit. Our dream and the goal is to scale this infrastructure to 50,000 to 100,000 nodes at a moment. We have run some of our technology test cases around this scenario and proven that the technology can scale up to 200,000 nodes. Centaurus Project is all about scalability, low latency, also it supports the cloud native infrastructure along with the VMware. The project Centaurus consists of two products, Arctos and Mizar, and you can find them in the community and more detail in the future speech from other speakers from our community. On the edge computing side, as everybody is aware of, Edge is becoming a necessity as day by day where the decision making and low latency and with the less network frequency, the computing has to make the decision. So there are small sites that are going to be there as edge node. The many applications that we are looking at is the railway application, the agriculture applications, sort of solar industry application, cryptocurrency. These are all related to the edge computing and when the edge collects the data, those data will be transformed to the private cloud or public cloud for the further processing. The last topic I am going to cover today is AI. The AI project in our world is called as FORNEX and you can also create the fork from GitHub for all these projects, Centaurus, Arctos and Mizar and FORNEX project and you can actually experiment and contribute to the community. Under the AI we have been developing some serious services that can work on the Centaurus core capability and will allow you to have a better decision makings based on the current scenario. As click to cloud we are also developing an AI based solution which can allow you to basically think more philosophically. In my world the story is when you are born you are only having a genetic information and beyond that everything else is a discovery. The moment you discover an information it becomes an historical information. Human intelligence is nothing but constant discovery and connecting it with the historical information that you have preserved. On the same concept we have wrote application called Cloud Intel in partnership with Centaurus applications. Cloud Intel has an ability to discover data center on premise or the cloud applications along with the licensing capabilities. And then we have Cloud's brain which runs on the Centaurus infrastructure which allows you to have a multi-cloud integration where you can actually monitor and store all the cloud activity on that system. So on the same concept of human brain Cloud Intel does the discovery and Cloud Brain makes the decision making. That's how the whole AI to the human intelligence is mapped into our system and to scale that system Centaurus is something that has been helping us to scale it to a massive amount of data as well as in the decision making capability. I thank you all for attending the session and we look forward to meeting you in Japan soon. Thank you. So as we all know that we are in open source and it cannot work without a community and the supporters like you all. This like I'll just give a quick background about what the Centaurus like we already know about this project, the technical concept but how it actually started and what are the key partners that has been involved with this community and what is our goal so far where we have been and what we want to achieve and without your support it again will not be possible. So as we discussed like Centaurus just launched in Linux Foundation officially and on December 16 2020 it's still a small baby but we are still like growing that and with the support of the great partners like Futureway, Click2Cloud, Great Green again, Soda Foundation, Informatics and a few others. So the partners got involved to discuss the strategy of course with the great community members and industry experts. We all know that there is a huge demand in industry when we talk about the edge or even the growing number of IoT devices that we use on daily basis at the same time the medical industry, healthcare or even a smart city like the devices or cameras that are based around the places and it required the mechanism to manage and to process the data. We found that there is still a huge requirement in industry at the same time there is a gap which needs to be fixed and that is how the Centaurus comes into the picture where we are trying to make sure that how we can make the distributed cloud at the same time provide a multi-tenancy scale like the architecture or the nodes at 30K even in future with the multiple numbers or thousands nodes. So our entire strategy comes into the picture that how we can make that possible. We form a community with the voting system or with the support of a different industry and leadership. We got an advisory committee which is part of Dr. Shong, Dr. Professor Hakim and Chris from CNCF where they are part on the advisory and then defining the technical or industry requirement even the strategic guidelines that how this community should look like. Then we got like entire technical steering committee of member of seven members elected by again a community and industry like Deepak, Prashant, Stefan, Sneel, Showning and Nikita. So we all are, I mean there's all technical steering committees there who meets every single month to define the project to make sure that what are the missing parts or even to add a new projects like approving the industry, approving the guidelines and making sure that we are meeting the industry expectations that is going to come in the upcoming quarter or even in the current phase. With that we got like outreach and marketing committee which part of Annie and she's not here today but she's also there and myself where we are taking care of the marketing partnership without that which is the core part along with the technical capability. So that is how our committee looks like and again along with this committee we have a special interest group like SIGs for each of the modules that Dr. Shong, Magni and Peng have just introduced like Actors, Mizar, The Edge and AI side and whenever we define this project since it's open source we all, we want you to be part of this community even this project and start developing or even giving the suggestions. So this all SIG groups are open source. I mean the meetings which we conduct on a weekly basis is open source. You're more than happy to participate and give the suggestions or a recommendation even if you want to suggest any project that you think is relevant that is where we normally discuss and communicate in this forum. So we have four different SIG groups. Then recently as I mentioned that since this community is, I mean a project is still so young but we got a, we recently hosted an event in Asia in APAC region last month where we got a great response from about 12 different countries. Like a lot of speakers joined, our industry experts joined from China, from India, from France, from the different regions and they were part of the keynotes. We successfully launched this event where the traction of that Centorious project went to 50,000 outrage on the social media channels. Around 1300, like 10 people registered for that event. 200 colleges in Asia Pacific showed interest to get that community project in this curriculum activity. Around 700 plus people joined live during that event. We got 40 members been part out of that event and then who might be a core community members even in future or can be a core contributor during this Centorious project. We announced seven top awards along with the cash prizes and 33 certificates been sponsored by our partner I mean Collector Cloud. So this is a public release and announcement that has been published on the local and regional newspapers are the gifts that we have announced or distributed as well as the vision to make this project to the industry like telecom to health care and so on. So that is how that vision has been recognized by Government of India as well which has been published on the announcement that you can see on right. So with that I'm sure that you might be interested to know more about how you can join this community and how you can start contributing on being part of this ecosystem. So you're more than welcome to join we have our booth entire four days in this event our team is there you can feel free to reach out and then talk about the ideas that you have or even the feedback in case you think or want to learn more about that. With that you can reach or go to our website and know more about the Centaurias we have our mailing group subscribe that mailing group so you can get new announcement and the latest updates that we have I mean released so far we have our GitHub accounts where you can fork the code try out and like do some kind of activities around it and feel free to share as the feedback we saw that we I mean even in that small duration there is thousand lines of code that has been updated in past couple of months from the community contributors like you all so we welcome all of you to join our group also the Slack for any offline communication we have different Slack channels for each of this project and with that meeting information can be also available over there so I with that I'll just let the time give back to you to enjoy your lunch and thank you so much for being part of Centaurias and we welcome you all to grow this community even more further thank you. Hi my name is Deepak Vej and I'm a cloud technology strategist at Futureware Cloud Lab in this talk I will present deep dive into Centaurus Centaurus is an open source project so project Centaurus as I mentioned is an open source Linux foundation project and this Centaurus platform is targeted towards building unified highly scalable public or private distributed cloud infrastructure so distributed cloud is thing to emphasize here and I'll come back to that later Centaurus cloud infrastructure project aims to meet the challenges for new types of workloads such as AI and 5G application landscape so these workloads typically exhibit characteristics such as mobility low latency location awareness and privacy and security and that's in order for to address all of these issues you need a different kind of cloud infrastructure which is distributed in nature so that's what Centaurus cloud infrastructure is it's a distributed cloud infrastructure platform to address all of these needs so Centaurus enables enterprises the hyperscale capabilities that slide towards dramatically changing the economics of IT so it essentially allows enterprise IT to build a hyperscaler cloud platform within their own environment now Centaurus is an overarching packaging just like OpenStrike is overarching packaging for underneath sub projects such as Nova and Neutron similarly Centaurus has key underlying technology as integrated sub projects first of these sub projects are the following Octos is a large scale cloud compute platform just like Nova is for OpenStrike Mitsar is a high scalable and high-performance cloud networking and Fornex is a autonomous and flexible edge computing sub project and Alnair is an intelligent platform for AI workloads so we'll talk about Octos compute project so Octos is an open source project designed for large scale cloud infrastructure and this was evolved from the upstream Kubernetes codebase and it features a lot of similar API objects if you're familiar with the Kubernetes you know, objects such as Spards and Uptica sets you know, we have the same concept as part of Octos just like Kubernetes but we had to do major surgery and enhancements to upstream Kubernetes in order to incorporate core design changes for enabling the following key features then these features are the unified cloud infrastructure resource support and we'll get into that what that means high throughput and low latency and multi-tenancy support so that's so this is kind of a high level pictorial architectural overview of how Octos looks like as you can see that we have partitioned the control plane at every level in order to achieve hyperscale level scalability this this required major surgery and key design changes to the upstream Kubernetes codebase so we introduce new constructs such as tenants tenant partition resource managers so these are all the new constructs introduced as part of the overarching Centaurus architecture so hyperscale so now first the key capability of Octos is the hyperscaler cloud scalability so public it enables public cloud level scalability it aims to support 300,000 nodes physical host per region and about 100,000 nodes per cluster within a region so you can see that this is a lot of host this is really how hyperscalers run their cloud basically and which you typically don't get that in upstream Kubernetes open stack environment and all the control plane components in Octos they can scale out independently and they are highly available so tenant workloads are partitioned so that allows extreme scalability the multi-tenancy so we had to do major surgery in order to enable hard multi-tenancy as a first-class citizen which basically allows multiple customer cloud cloud instances to secure securely coexist in a single physical octos cluster so essentially you can have multiple tenants in the same physical cluster with the hard multi-tenancy and one tenant is totally unaware of the other tenant in the same physical cluster so in order for us to do that this required allowing tenant ID to flow through the whole entire Kubernetes code base so that includes namespace aware API objects non-namespace API objects as well so the unified runtime orchestration so the contemporary fragmented orchestration stacks for containers and VMs containers for example Kubernetes and VMs open stack they introduce a lot of issues issues such as resource pool inefficiencies duplicated components orchestration costs so you need one stack for orchestrating your VMs and you need another stack to orchestrate your containers so that leads to all of these inefficiencies so what octos does is octos introduces the native support of VM in addition to the mature container support inherited from the upstream Kubernetes so by doing that it provides a unified resource resource pool so it provides abstraction runtime abstraction whereby it not only currently supports VM and containers but in down the road in future it can support other workloads such as WASM, WASI, WebAssembly or Unicarnals as new runtimes as well so it's a plug and play actually so currently we have enabled the VM in addition to the container networking as part of as part of octos so the next thing we are going to cover is a META network so META just kind of a high level 30,000 feet level definition META is the network virtualization layer for our next generation Centaurus cloud platform so large scale cloud platform like Centaurus they need to scale in order to support enterprises global footprint just like hyperscalers do actually so we wanted to and the key thing is we wanted to support rapid provisioning of cloud resources very quickly short span of lifetime for workloads such as containers and serverless functions so from days and hours to minutes you know and fraction of a second for dynamic for such a dynamic cloud environment so it's not your old way of doing your cloud networking where you maybe spin up 1000 VMs a day in a dynamic cloud in a hyperscaler cloud environment you can spin up millions of endpoints containers or serverless so in order for you to be able to do that you need a very rapid provisioning support so none of the contemporary open source networking solution can measure up to such a large scale cloud requirement so that's what we ended up doing we ended up building META virtualization layer from ground up so the problem with the programmers thinking in flow rules so currently in an OVS space typically if you're building a cloud environment you would use flow table flow rule space OVS kind of environment you see in a typical such an environment a control plane typically programs flow tables on each of the virtual switches on each hose using the open flow protocol so the current flow based programming solutions are not scalable and they have lots of issues and quirks time to provision ports increases significantly at the number of ports high CPU utilization during flow passing packets travels multiple network stacks on the same hose provisioning time of a new workload depends on the number of workloads already existing in the system so if you adding a new endpoint to a subnet which spans across multiple hose you have to go each and every hose to kind of update the flow rules for the new endpoint so you can see that unfortunately such an architecture doesn't meet the scaling goals of Centaurus hyperscaler great platform so to address the scaling challenges we built from ground up what we call the META networking there virtualization there in order to route traffic for virtual network so why did we have to kind of build the META networking there from scratch so the ongoing state of the art work there's a lot of good work going on OVS, OVN folks and the Andromeda project but all of these efforts were still trying to kind of mitigate the issues related to flow rules based system data plane so they're trying to reduce the number of flows passed down to the data plane so to address the challenges which we talked about previously so instead of doing that what we did was we took a clean slate approach so we build our networking network virtualization there the host data plane there as a regular distributed system application so instead of age old networking design constructs such as mac learning, flooding the tunnel and etc none of that actually so we built our network virtualization just like you would typically build a distributed system application and we'll get into that how we did that so before we get into what META is so META essentially is built on top of XTP based XTP stand for express data path it's a Linux kernel capability so essentially it's a whole space programmable data plane abstraction provided by Linux kernel and XTP is a Linux kernel hook for running an eBPF program within the device driver of the META and then just kind of a high level a very brief description of what eBPF is eBPF is a sandbox program that runs in the Linux kernel without changing the Linux kernel so it's a very it's a very small 4K instruction size and written in any generic language mostly in C but you can write that in Rust as well and all of these eBPF programs are verifiable programs essentially before loading into the kernel the verifier verifies that there's no loops there's no wrong pointers or memory allocation and things like that so essentially what happens when the packet arrives in the NIC you run an imperative very high level logic using typical data such as arrays, hash tables and then make appropriate action to the packets those actions are you either pass them to the networking subsystem within the kernel or you transmit them back to the NIC or even redirect them to another interface or drop them so these are the four typical packet actions you take but you do that in an imperative manner as opposed to the flow rule based declarative approach so this gives you a lot of kind of an advantage actually to be able to imperative approach is much more powerful as opposed to doing your network programming using a declarative flow rule based approach so this is what the high level architecture is what we did as I mentioned before so we removed everything from the holes such as OVS, OpenVswitch, Linux Bridges, IP tables, none of that actually and what we have is XTP program running on a main interface we call it a transit XTP that's the one you see there at the bottom and then we have another program of the VF pair that connects to an endpoint the endpoint could be a container or a VM it doesn't matter so it's a unified networking environment and we call that program XTP program as transit agents and these programs share EBPF maps which are programmable from the user space by the transit demon as you can see that at the top so on the management plane if you would like to push configuration just make a RPC call to a user space program and these user space program populate regular maps hash tables inside the kernel and all the XTP programs operate on these maps to perform all the packet processing so the extensibility so the XTP programs are very extendable one of the key characteristics of XTP be able to do a take-all so instead of having a one XTP program so we have multiple XTP programs attached to the next issue so essentially what we have is a one XTP program which is called a primary XTP program and depending on certain matching conditions you may spawn another XTP program by doing a tail call to it so that makes this very extensible so it opens up a lot of opportunities for you to kind of build this extensible networking virtualization so in summary so the floor programming model is great for programmable switches but not scalable for multi tenant virtualized cloud networks and it's tremendous provisioning through port and runtime CPU performance gains it creates an extensible plug-in framework for cloud networking and unifies the network data plane for VMs containers other workload types so from a networking standpoint it doesn't matter if it's VM containers or serverless and other capabilities such as label-based network policy enhancements so and one of the key thing is programming the be able to offload the XTP logic to the smart take it's pretty attractive as well actually so and so that's another thing that was very we looked at it from the beginning that the XTP's ability of be able to offload it to a smart take allows us to kind of have offload all the our infrastructure for to the smart take and then perform all the network packet processing logic there so that's pretty much covers meets our networking and the next project is for next edge computing so this is so for next is an open source edge computing framework for managing computer resources resources on the edge and this is a project is designed to solve some of the key edge computing challenges such as limited computing resources heterogeneous resource types topology unreliable network and long latency so with the for next end users edge application workloads to be easily deployed in a distributed hierarchical edge environment with topology is that best matches the physical and logical structure and it offers high performance high virtualized virtualized networking for workload workloads communicating within and between the edge cluster so it's not only not sour but be able to communicate east-west across clusters as well so these are the key features provided by for next project so computing nodes and clusters on the edge so both computing nodes and full-fledged clusters can run on the edge we'll talk about that I have a detailed slide next slide and hierarchical be able to support hierarchical topologies and then flexible so you can not only in a hierarchical topology you can not only nest your octose cluster but you can not nest your Kubernetes cluster or K3S as well and that's networking be able to do multi-tenant edge cluster networking and supporting you know the constructs such as VPC subnet and high performance inter-cluster communication the east-west communication whereby the edge nodes they can directly talk to each other without going through the center cloud which is a requirement because the low latency is highly desired at the edge cloud environment so this is what the hierarchical topology looks like you can see that you have a center cloud typically you're familiar with the cube edge project you know center cloud manages edge nodes but here what in the forenext project we have done that is we have gone beyond that so you not only have your center cloud manage your edge nodes but it can manage a full-fledged cluster at the edge and those clusters in turn can have sub-clusters sub-hierarchies basically so you can have a nested hierarchy so essentially so this is what I meant by the hierarchical support for hierarchical topology so essentially when you provisioning a workload you can go to the center cloud and then depending on your placement policies your workload will get deployed on the appropriate edge environment basically just a simple edge node or it could be a full-fledged cluster itself Elnir is our AI project so essentially the vision for Elnir project is building an intelligent platform to improve AI workload efficiency so AI workloads will be critical dominant workload for the cloud and edge computing we talked about the emerging 5G and AI application landscape so the current cloud and edge systems leverage existing hardware software architecture to support new AI workloads which limits the capability of AI training inferencing and also increases the model serving cost as well so one of our vision is more efficient and more intelligent hardware software frameworks and architectures in order to support AI workloads and focus on the resources resource management aspect to analyze and schedule AI workloads on existing new systems with intelligent methods we also exploring new architectures to orchestrate heterogeneous resources and new service model to facilitate AI workloads so these are the key features provided by Elnir so Elastic platform training capability Elastic training dynamic GPU allocation GPU utilization profiling precise resource management GPU fine-tune fine-grained sharing optimized resource utilization autonomous scheduler continuous scheduler decision learning policy improvement and optimized ML framework parallelism at the data parallelism model and pipeline optimization and hyperparameters auto-tune that bunch of other they were doing very interesting inferencing ML inferencing project as well collaboration with University of Washington as well so this is this pretty much concludes the stock so and this is and I can this is a half an hour slot this is the best I could cover actually the tons and tons of information available a lot of documentation on our website and our GitHub as well and it is it's a pretty active project and there's the tons of work going on in these projects and we'll be very happy to have the community get involved in all of these efforts you know the meets our networking octos compute computing and L layer AI so there's a lot of work going on and we'll be very happy to have you folks actually get involved in these projects so with that you know this concludes the presentation of Centaurus distributed cloud infrastructure deep dive