 Okay, hello everybody and welcome to this OpenStack Newton on OCP hardware presentation, where we will go through how perfectly these different open source projects fit together. My name is Juha Ravainen, I'm working for Nokia and I'm coming from mobile networks architecture and technology unit. Hi, my name is Juha Kosonen, I work as a software architect on the same organization as you. Agenda of our presentation is such that Juha, okay Juha and Juha, so Juha Kosonen. First go through this OCP and what OCP is all about. Then I will cover this OPNV part and I'm telling you what is OPNV and what is OpenStack's role in OPNV. And then we will go through what kind of tests have been made with open source software on top of open source hardware. And at the end of our presentation we have a short summary. Okay Juha, please go on. Thank you. So a little bit of OCP history for a start. In 2009 Facebook was offering new services and giving millions of people a platform for example to share photos and videos. They realized that it had to rethink its infrastructure to adapt the huge influx of new users and data and also control costs and energy consumption. To accomplish this they started a project to design a world's most energy efficient data center that could cope these challenges at the lowest possible cost. In 2011 Facebook shared its designs with the public and along with a couple of other companies launched the Open Compute Project and incorporated the Open Compute Project Foundation. The aim was to create a movement in the hardware space that would bring about the same kind of creativity and collaboration we see in open source software world today. And this is something which is actually happening. The specification and designs are contributed to the Open OCP foundation through the OCP projects. There are several projects such as server, networking, storage and also telecom among others. So the put it short basically the Open Compute Project is redesigning hardware and making it more efficient, flexible and scalable. Here is the list of OCP patinium members or at least this was the situation last week. There are many other companies on other membership levels so this is just to illustrate that there are members from different areas of the industry. For example ATT, Goldman Sachs, NVIDIA, Nokia and so on. Nokia joined the project in November 2015. And why OCP hardware? Well, OCP hardware is cost efficient. It lowers the cost of post infrastructure as well as operational expenses too. It's a vanity free design without unnecessary components and materials. Disaggregated rack level solution with common resources like power. Let's see the building blocks a little bit later. From maintenance perspective a major benefit is the fact that all hardware operations are easily performed from the front of rack. There is no need to access a rear hot eyes and no special tools required for these operations. Next we have a short video demonstrating the process of replacing a faulty server and the time required for it. This is a comparison between OCP hardware and rack mount. The procedure is simply to unplug the cables, remove the faulty server, install a new one and then plug cables back again. In the video on the left hand side is the OCP on the right. At first please pay attention what's going on the right hand side as there the work is finished already in the quite early phase. So let's put the video rolling. So pretty nice. This was only a very simplified example of one unit replacement in a single rack environment. So you can imagine what would be the case in a real data center. And then some more benefits of OCP. The high rack efficiency. OCP 48 servers per rack mount 32. This means a remarkable improvement in rack footprint per compute node. And what comes to the energy efficiency of OCP data center excels also there. And just a quick overview of building blocks. These are from Nokia airframe portfolio. So there's the rack. It's similar as 19 inch rack from the width point of view. Then there is power self with six power supply units, compute self or three computes, server nodes, storage nodes and switches. Originally OCP is not designed to fulfill regulatory requirements from CE and FCC. Nokia airframe include several enhancements, such as electromagnetic shielded doors, side panels and cable openings, improved seismic tolerance and minus 48 volt DC power feed, since that's what Delco equipments is traditionally using. So Nokia airframe is a Delco great solution. In general OCP has gained a lot of interest in industry. Some of the major players such as ADT is using Nokia airframe in OP NFV evaluation, which brings us to the next topic where Juhat tells what OP NFV is and how it connects to the open stack. Okay, thank you Juhat. So OP NFV is open platform for network functions virtualization. This project was established about two and a half years ago and it's a Linux foundation open source project. And through system level integration deployment and testing, OP NFV creates a reference in the platform to accelerate the transformation of enterprise and service provider networks. So all in all OP NFV is a reference open source platform. Scope of OP NFV, originally the scope was virtualization layer and virtualized infrastructure managers. So on the right hand side of this slide you can see those low layers, NFVI and WIM. So those were there in the scope originally and NFVI, it means network function virtualization infrastructure and WIM is this virtualized infrastructure manager. But later on also MANO was taken into OP NFV scope, so MANO is management and orchestration. And here on the right hand side you can see those VNF manager and NFVI orchestrator components. So the current OP NFV scope is inside of that red dot line. And what are the key components of OP NFV? I would say that the open stack is the fact of virtualized infrastructure manager component at the moment. But whenever you make this OP NFV software build, then of course you need also some other open source components. And actually you have several options how to build this OP NFV software build and you can choose different kind of open source components. And in OP NFV we are talking about scenarios. So there are quite many scenarios in OP NFV. So in addition of this open stack, if you want to take for example HDN controller into this OP NFV software build, you can choose if you want to take open daylight or own up or open contra. And then of course you need also Linux and you can use for example CentOS or Ubuntu or maybe something else. And to mention some other components, there is for example this open V switch so that you can take only the open V switch or you can take also with DPDK. So there are different kind of options how to build this OP NFV software build. And in OP NFV we have several installers which installs the build in the way you like. And in the latest OP NFV Danube release also the first mono functionalities came into OP NFV. OP NFV member companies so there are different kind of membership levels and there are a lot of companies in OP NFV nowadays. And just to mention a few of them you can see that there are big operators like AT&T Dynamo, for example there are vendors like Nokia, Huawei, Ericsson and so on. And in addition of that there are also some other big players like Intel and Redhand among others. So all in all there are now over 50 member companies in OP NFV and OP NFV is a quite big open source community nowadays. OP NFV has a six months release cadence and these releases are named after reverse and they follow alphabetical order so Arno was the first OP NFV release then came Brahmaputra and Colorado and the latest one is Danube and this Danube was released about one month ago and everything works has just been started. So there are four releases so far and more is coming. And if you wonder which OP stack versions are used in this OP NFV release so then I can tell you that Liberty was used in Brahmaputra, Mitaka was used in Colorado and Newton is now in this latest Danube release. Okata was actually already available when this Danube was released but Okata came out so short time before Danube that it was not possible to take Okata anymore into use. And this release work has just been started and OP stack Okata or bike will be used in E-release but I'm not sure because this work has just been started that whether it's been decided which one of those two OP stack versions will be used in OP NFV E-release. But all in all there is a new OP stack version in every new OP NFV release. And here you can see the OP NFV release history so Arna was the first OP NFV release and it came out a little bit less than two years ago and it was the baseline foundation of components necessary to build an NFV platform from upstream components. And then during 2016 OP NFV Brahmaputra and Colorado were released and the latest OP NFV release is Danube which brings together end to end networking stacks including management and orchestration, data plane acceleration and architecture advancements. And if we take a little bit more detail this Danube so that what now came out so like mentioned already a couple of times so this management and orchestration is now also in the scope Open O project brought some monofunctionalities to OP NFV. Okay, there has been some changes lately so this OWNAP project was established lately and this OWNAP is Open Network Automization Platform and it is the latest open source monoproject and it actually put together Open O and E-comp open source projects. Then there are some architectural improvements like network control flexibility, high availability and multi-site improvements. And in NFV performance area there are some data plane acceleration and for example this FDIO has nowadays its OP NFV project and in this OVSD, PDK and KVM or those components are naturally improving all the time. Then there are some feature improvements and like mentioned here IPv6 and SFC and layer 213, VPN fault management and so on and this OP NFV supports multiple hardware architecture so you can run it on top of x86 hardware or ARM hardware and it works also on top of this OCP open source hardware. Then one other thing which I like to mention from OP NFV so that in OP NFV we have these public labs so on the right upper hand side corner you can see that there are these public OP NFV laboratories around the globe hosted by different OP NFV member companies and these laboratories are meant for CI and CD purposes and testing of this OP NFV platform. And Faros project builds several different kind of scan areas and deploys them across the globe on multiple hardware platforms and if you wonder what this Faros is, so Faros is an OP NFV project and it's OP NFV laboratory infrastructure project which develops OP NFV labs infrastructure that is geographically and technically diverse and Nokia is also having one OP NFV public laboratory it's in Espo Finland and currently it's a Rachmont hardware based laboratory but we have a plan that we will expand OP NFV public labs with OCP based hardware in the near future and last but not least from this OP NFV point of view is this cross CI XCI so this XCI is a service provided to OP NFV community and the community's OP NFV work to enable early integration faster feedback alignment and reuse so if for example in open stack somebody makes a patch so then there is own testing pipe in open stack side and all those test phases needs to be passed so that this patch will be taken into the software build but this cross CI system it expands this methodology to all this testing phases so that OP NFV can take the latest open stack version from the main branch and then it tests how this latest open stack version works together with all these other open source components and if there is some bugs or some interoperability problems so then the feedback loop comes much faster so the idea is to discover these bugs in OP NFV environment faster and fix them and there are some plans also with some other open source communities and then we go to the testing part Yes, let's start with the discussion with the event called OP NFV blockfest blockfest is an event organized twice a year after a new OP NFV release it focuses areas like OP NFV deployment release, network integration, VNFs and generally tries to accelerate the open technology development and commercial adoption the second OP NFV blockfest was hosted by University of New Hampshire interoperability laboratory in Durham and Nokia participated this event too and actually this was the first time OCP was used in blockfest so there were 70 people from 23 different companies on organization attending to this event and the event itself is not only for OP NFV members but is open for non-members too and in this blockfest Nokia's main focus was in installer and a doubtful testing in OP NFV there are several installers that can be used for doing the actual deployment blockfest provides an opportunity for installer projects to run deployment against different sets of hardware in that sense running deployments on unfamiliar hardware helps to validate that design of each project is portable across multiple vendors hardware platforms and similarly this enables also hardware vendors to test and validate the compatibility of OP NFV software deployments on the products the deploy and test activities allow many participants to get hands-on experience with OP NFV software as well and about this dovetail testing so what does this actually mean in this context OP NFV port of directors has initiated compliance verification program which is known as CVP the program's objective is to enhance interoperability and reduce adoption risk and testing cost for end users by verifying hardware and software platform interfaces and components the verified products may have different features and capabilities but they are all expected to have a compliant by implementing exactly defined interfaces, behaviors and key features dovetail is the project that delivers open source software tool chain documentation and test suite for this CVP and it leverages many OP NFV and upstream projects to accomplish this the OP NFV certification and compliance committee can award the product OP NFV ready status if product passes the compliance verification shoot this verification process itself is not yet effective but we are expecting this to happen later in this year and this is the airframe setup we had there on site on the plugfest it's this sort of half rack size there are six server nodes and data management switches and power cells so it's the very same building blocks we saw already earlier and all in all Nokia airframe got a lot of interest from participants in this event since it was nice to have the hardware physically located in the same space where people was working so everybody had a chance to really see the setup and the plugfest report is publicly available on opnfv.org test activities, bugs discovered and reported to OP NFV project teams participants lists and things like that that they are all described on the report and the next plugfest was in April after OP NFV Danube release this was organized by Orange in their premises in Châtilion near Paris so big thanks to Orange for making this true this time topics such as DPDK and INIX were our team's main interest and we had a lot of good cooperation and with many people especially on issues related to these topics among other topics of course an event like this provides a good opportunity to meet face to face experts from different companies and organizations to collaborate and utilize their expertise on the work currently in hand and this Danube plugfest report is not out yet and then few words about VNF testing in Nokia internal OP NFV test lab so the VNF means this virtualized network function which basically you can think with the VNF as a real network element for example a base station deployed on the network function virtualization infrastructure in Nokia OP NFV test labs we are running our commercial VNFs on top of OP NFV using different hardware and also commercial NFV platforms and by using a different combination of hardware and platform in testing helps us to verify the capabilities of underlying platform and to identify the potential effects it may have on VNF site so and next short recap by Johan Okay, thanks. So Chammari and then so this open source software and open source hardware is now leading us to an open future so this open source ecosystem it is the way to accelerate the introduction of new NFV products and services Traditionally open source projects have been working mainly inside their own communities but nowadays cooperation between these open source communities is increasing all the time like we have described in this presentation so when those service providers, enterprises and individuals are all consumers of open source but at the same time they can also be contributors to open source projects which is great and we already mentioned this OP NFV plug-fest in this presentation and OP NFV plug-fest are a great place to collaborate and I suggest you all to read these plug-fest reports plug-fest are organized twice a year and the latest plug-fest was organized couple of weeks ago in Paris report is not yet ready but it will come out in the near future and actually these two sentences are directly from this latest plug-fest report and there you can see that many OP NFV community members were able to deploy and this OP NFV for the first time with this OCP based hardware and this OCP gained a lot of attention in this event and the community is looking forward more collaboration opportunities between these open source communities and in open source stakeholders are CSPs, vendors and also individuals and openness leads to faster innovation cycles and lower vendor lock-in bigger, better and faster results can be achieved with shared innovation research and development open source allows shared R&D between CSPs and vendors and nowadays also customers can create new features directly in open source communities by contributing so instead of creating a feature request towards vendors customers can also contribute directly to open source projects in open source global communities come together around a shared purpose and communities are able to create things that go behind the capabilities of any person so we have told you some examples from the cooperation between different open source projects and it looks like the future is built on openness and open source is one important part in that new technology will leverage open source right from the beginning various current networking technology evolves to a cloud native future that also builds heavily on open source and more is needed in the future factory robots will communicate wirelessly and cars on the highways will exchange the information with each other this requires extremely low latency mobile networks known as 5G the first cloud networks have already been deployed but more is needed with future networks as more functionalities will be moved to the cloud so I would say that the future looks interesting and even more cooperation is needed and this was the end of our presentation and we want to thank you from your time and now it's time for questions Estimate the penetration of the OCP platforms in data centers and NFE specialized data centers How many data centers are out there that are actually running OCP hardware versus common off the shelf One use servers, other module, modular platforms I would say that this course may be a little bit out of the scope of our presentation so we just told you what kind of cooperation we have been made with all these open source communities so I don't have the market figures so that I could tell you how popular this OCP hardware is at the moment so sorry I can't answer this kind of question I have a question, compared to a similarly equipped say one use server you mentioned in earlier presentation that it will cost advantages to running OCP hardware I would then expect that there will be like an OCP board would be 30% cheaper, extra percent cheaper than one use server I don't really find that to be the case, can you comment on that? Can I comment the density? So I think maybe the density is one issue so that you can get the more of these compute nodes power in the same footprint and so if I understood the question correctly so you mean that the one server is not cheaper and maybe that's the case but when you make big data centers so then you can gain the benefits there I have a question related to the architecture, a roadmap Do you have any update or any insight in terms of something similar to rock scale distributed architecture in terms of OCP? Rock scale architecture, well I would say that this goes also a little bit out of scope because the main idea of this presentation was this cooperation and so there is also OCP stand in the marketplace so if you want to discuss with those guys but then this other technology so I can help you with that unfortunately Okay, I'll obviously no other questions so thank you for my time