 All right, sounds like we're live welcome everybody Back from lunch. This panel is on 5g glimps into a massively connected world I've got three very distinguished panelists with us today, and I will let them introduce themselves Starting on my left with Heather. Hi. I'm Heather Kirksey I'm the VP of NFV for the Linux Foundation, and I head up the opnfv project Which for those of you who may not be familiar with it is an open-source project focused on integrating deploying testing The next gen software based networking stacks Hi, this is Balaji ethirogyla. I work in Ericsson as a director product marketing This is my favorite subject 5g network slicing management orchestration who will enjoy the show Thanks Good afternoon everyone. My name is Ling Li. I'm from China Mobile I previously work for the open all project and now on the TSE of on app and actually internally I'm part of the Nova net project inside China Mobile which drives the strategy and also the vision of the next generation network for the company And we are actually doing both open source projects and internally with various trials So and we look forward to bringing new technologies into trials and also quick commercial adoption Very good Thank you. You are. Oh, that's a really good point Hi, everybody. I'm Phil Rob. I'm the executive director of the open daylight project as well as the vice president of Operations for the networking and orchestration projects within the Linux Foundation. I Spend most of my time right now pretty much splitting between the open daylight and own app Balaji, let's start with you for questions just to give us a background What are the key technical attributes of 5g infrastructure band width device density and so forth? Are we gonna look forward to in 5g? Thank you Phil Yes before I talk about this particular, you know the key attributes of 5g Let's look at What are all the operators challenges as they think about 5g or migrating from 4g to 5g? How 5g will address those challenges, which is key attributes actually? I will sum up the challenges into three categories main categories the very first one is Operators are expected users are expecting really complex services, especially in the area of Multiple vertical industries so one service will pass through many vertical industry before being consumed by the end users operate That's a big challenge complex services and the next is as part of the 5g I mean we have that in some sense in 4g because of virtualization, but 5g People will expect operators and the users will expect even much more advanced automation in agility So what is agility in terms of in the context of 5g? So if you see agility for example in general terms when I receive a call from my enterprise customer or from my user The way I take the requirements and the request and process it quick and create a new service, right? That's basically time to market that has to be really really fast cannot take three months Sometimes it may be expected in 15 minutes in 5g area. The next is fulfillment Okay, we defend the service how fast you can fulfill and activate the service So the agility is key and that can be only delivered if your network is highly programmable and Modular that's very key 5g will bring all of that in the last but not the least you probably know most of us cost So they want complex services to be delivered fast and automated way But at the yet at a very cheap price, right? So cost is a key part. So 5g will look at all of these things To you know, some of these are the key attributes now going back. What is 5g? Right, that's a critical to understand because we talked about 1g 2g and 4g. So 5g is not just purely Just the radio technology 5g will include Basically different access the transport as well as the cloud in terms of number of technology Which is 5g based on as of today 5g actually will take even further new technology that is coming up but as of now it will include network virtualization network functions and Automation management and orchestration and many of this will come together and actually will also for security Blockchain is coming up. So we'll use for blockchain for data integrity So all of this will play all these attributes will play a major role in delivering services under 5g Thank you Anyone else want to comment? If I may to add I think as Bellagio is just summarized You know for 5g absolutely, you know, it puts a very peculiar requirements to the network and some of the you know for carriers traditionally we we provide Very few types of services and our network is actually cultivated to that small set of Services, but for 5g You know, especially we have very different applications coming coming in and it has to be deployed and managed and in an eye-draw and Reliable and also for Separated manner on the same infrastructure. That is the vision of 5g and so as administration and automation is absolutely a key area a requirement to that to enable agility, but also we believe that Industry standards and comments like consensus of how vendors would implement Different like services of 5g. For example, there has been discussions about microservice architecture of a 5g applications and how that would be implemented in a network function granularity or in even lower a Final grand like microservice on level and how vendors could work together to put Put together that into our feasible solution is still a big question And one remaining I think is one of the challenges Very good very good And I know certainly from my perspective just as a user as we've gone from 1g 2g 3g 4g Now moving into 5g we can certainly expect through the radio technologies, you know significantly higher bandwidth To support better video and so forth through our wireless networks But also with the expectation expectation of iot and the explosion of endpoint devices It's really also about the defense the density of those devices Just in general from a cell perspective as well as certainly from my You know intense densities from an urban setting So starting with that Those are some of the requirements for 5g and then of course there's the one millisecond Latency Expectations so we've got applications that are going to require much faster response times And as such a lot more changes than just the radio frequency to support 5g Lately so can you tell us a little bit about some of those new applications that are to be enabled with this technology? Through that improved infrastructure Yes from our view on to start with Actually have these enhanced mobile broadband applications families including virtual reality and augmented reality which would require you know Also, I think the ultra latency requirements as you just mentioned and the other type is I think you know the massive iot applications for example vehicle to X and also the smart factory Applications which also I think also require out to reality reliability and also low latency and Some of the other applications like traditional on core network and transport services and also we wanted to enhance the network architecture and also Some of the more I think features unlike Agility to deploy more on reach applications while these common infrastructure and and new technologies Just add on to what lingley has just mentioned One of the key Part as part of the fight again virtualization has has already given us that opportunity right that ability to do things So for example, we're talking about distributed data centers Centralized was as a distributor. So some of the radio functions can be Centralized but in a distributed data centers and some of the core functions. They they actually live in Centralized data center. We can move some of them much closer Especially the one that is related with the user plane for example media codex Video and all these codex can be moved much much closer to the user so that the data plane stays as close to the user as possible But going back to some of the use cases is for example in 5g We can expect You know fiber fiber speed actually so it may actually work for fixed network as well So some of the alternatives towards instead of fiber we could leverage 5g and also Real-time Applications for the car for example just give a typical idea if you use the same principle in 4g for example Use the car analogy and if you want to apply break for the car It'll take using 4g network because of the latency. It could take somewhere around 4.5 feet To make a decision to apply the break But in 5g because of this extreme stringent requirements on the latency We could do that that in couple of inches So that's what the expectations so it will tremendously open up new opportunity for the new you know Different industry take an example on robotics the very expensive machines you mentioned about the factory is exactly smart factory so In robots for example You could take the control function the brain and take them and put them in a distributed data center Closer to the production environment say if you have 50 robots They're going to cost a lot of money But the brain power is the key one that we take it and put it in the distributed data center But you need to have a less latency The good thing is you don't need to have a 50 brains could have 30 because it's virtualized We'll never in all of this you know the virtualization and the SDN will give those capabilities Yeah I mean when I think about the applications I mean they they've mentioned a lot of the ones that you know I hear on people's minds what I usually like to think about are the ones we haven't invented yet Right, I mean you know the movement from when we went from your old flip phones to our smartphones and all of a sudden the world was different So you know when I think about things like NFV and as dn and their importance for the network it really is going away from these sort of Special purpose boxes to this idea that you've got a flexible Upgradable DevOps agile infrastructure so that when all these crazy new ideas come out Your network is prepared to be there for you Right right and for me that's that's a lot of the tie-in in the sense that You know this is to support 5g We're going to have to move compute and storage much closer to the edge Which means a lot more management and orchestration control from within the networks where it used to be pretty much a centralized cloud That leads back into network function virtualization and the activity being done by opn of v So Heather you know as as you look at that How does opn of v fit into 5g from your perspective? How are the service providers using OpenStack and opn of v to build the software infrastructure needed to enable those new classes of application? Yeah, and it's kind of what I was saying before right you know where we're Reimagining the network instead of being a set of boxes that do things to a flexible software infrastructure And so which is basically the promise of NFP So to the extent that you want your next generation mobile network to have that flexibility to have that agility You know NFP is really at the core of that You know and what we've been doing with OpenStack is you know if you start building up that network There are a lot of pieces that are necessary. So OpenStack Is very important to us also your SDN controller There are various types of data plane acceleration technology that will certainly be important in a lot of these high-throughput Applications analytics that you might want to do on top of the network and so within opn of v We you know basically sort of work on the integration of all of those and make sure end-to-end use cases work across The different pieces that everyone is building necessary for the network If I may add absolutely Open source will play a major role across up and down the stack So like I said 5g is not just one area of technology just to give a typical example We work with Fido a lot, which is data plane, right? So universal data plane highly accelerated because packets going in packets coming out That's the basic thing right, but it needs a very low latency and also we don't lose the packets Right. So for example, just give idea here. We have an open daylight. We have a project called network That will integrate with Fido using any comb. This is a very typical example But small thing but if you see the larger scheme of things in 5g everything matters all these things are integrated and connected What I also think is going to be really really important going forward is you know We're here we're at open stack where to halten probably towards the you know technology side of the world But just you know sort of cultural and process wise when you get to start thinking about a more flexible network or more software-based network What that does open to you But it does change when you're thinking about management and things that it's you're not managing a bunch of boxes you you're managing compute and storage and network infrastructure and so you know a lot of this sort of thought processes the NFV sort of requires but enables I think are still you know We're still kind of figuring out how to wrap our heads around thinking about the network in a different way But I think that will definitely be key to really take advantage. I think of what 5g might be able to enable So I'd like to open it up for questions from the audience as well We've got mics here on either corner or in either aisle here So feel free to stand up and and ask any questions that you might have All right, but we'll continue on. Oh, no, we do have questions. Excellent. Yes, sir. It's almost on Tap on but he says it's on try again Nope There we go. Okay. Yes, we can hear you now. Awesome. Um, so this is a question for you Heather Um, you were talking about like a performance in NFV. So obviously the 800 pound gorilla has been We heard that So I'm just kind of curious some working specifically in the V space We haven't seen really any big integration as far as the data plane side and high-performance IO Inside of containers and I'm curious if any of the four of you have been looking at that or doing any development in that space And what we might expect Yeah, that's certainly a big topic on Everyone's minds right now sort of how Containerization and Kubernetes sort of fits into the stack that we've been kind of working with for about the past two and a half years You know, I think China mobile actually sort of just launched a project in opening a fee specifically looking at container requirements We're basically we're beginning to to figure that out And I think that there's still a lot of open questions about the best way to do it and what sort of workloads will work best You know with what mix of technologies also depending on where they're placed. So I Think the answers are the questions are just now being asked and the answers are being figured out Is what I'm seeing at least It's a very good question on containers a couple of things on the Fido side We're looking at you know the containers version of Fido and also integration of Kubernetes That's on the open source side, but from a vendor perspective. I could add from you know, large vendor in a 5g core Many of you know, we're right now. We started in 2014 actually virtualizing So the journey has already started before we talked about containers, right? So since 2014 we started moving our operators started asking as we started moving the function actual network function to the Virtualization right virtual network functions VNF Traditionally it was in VMs, but now we are going towards microservices. What we call now is a cloud optimized VNF So it is a deal, right? You just cannot look at web scale. Yeah, I'm going to just do that Telecom is a little bit different because of the many real-time requirements You can't shut the phone off just like that. It has to work 24 by 7 so there are very stringent requirements So that's what we're talking about cloud optimized VNF. It'll be microservices based and Containerized, but most important thing is there are some design principles. We'll apply that will be very telecom great And one of the complications to write is for most telecom network-centered applications There's a great deal more state to manage So figuring out how that works in you know some cloud native, you know concepts is also something that You know needs to get figured out people people were beginning to figure it out And I have no doubt that our community will but they're just you know There's some some you have mismatches of sort of points of view kind of coming towards that that'll be it'll be fun Figure out Fee has been doing great job for the system integration verification finding the issues feeding back to the community Developing it awesome and Erickson is the pioneer for the radio and working with open NFV together And now with Erickson Cisco right out for the FTIO with the honeycomb integration of their perfect job The only thing that I see missing which has been tackled by the court people is the network slicing So what is your view to implement network slicing either part of open NFV or totally keep it out or Let the others to figure out how to handle it for repurposing the whole network by means of slicing deslicing And what's Erickson's view to contribute in that work? Thank you You can take it I am sure that you have some views on network slicing, so I'll let you thank you for that. Yeah Yeah, so just address the concept of network slicing is very critical the piece so Okay, you have the network and the slicing will include Obviously the radio piece is important also the transport piece as well as the course piece. I mean Core space right, but it's beyond not just the network itself Network slicing also involves service slicing very important So there are I see three layers of orchestration playing together to deliver the network slice One is at the infrastructure cloud infrastructure layer, which is heat orchestrator in this case if you take open stack Or it could be a weak load director whatever and then that gives the flexibility on that layer That's very critical then take up the next layer Which is a network management or a VNF application level orchestration for example NFEO all of that So you create more flexibility there in the last piece is a service layer. So that's more of a business layer So, you know different businesses. That's what you create the service So we provide three layers of flexibility to create the network slicing right so network slicing that means not just you know your You know core network. It should be from the radio. Did I answer your question or yes And no my question mainly from open source of our community perspective It's driven by the court people and court especially but nothing from open NFV to drive this to become a reality So what will be we will be seeing as reality implemented by Ericsson driven by the open source Yeah, so yeah pretty much for our for our first two years You know we were we were fairly data center centric You know kind of you know in sort of what we were trying to pull together within opnfv We weren't really looking at at the multi-axis edge quite as much We've actually just started some projects around that both sort of with you know integration of kind of the more data Center stack with core potentially at the edge also open daylight has been doing a lot of work around sort of getting edge access and in the you know the integration with the optical equipment and the line and speed Leaf and spine and all that kind of stuff. So we're we're just we're basically bringing those in sort of right now And looking ahead to see some of that in our next Euphrates release So basically I think So to answer your question There's been ongoing work within folks like open daylight that opnfv is now I think ready to start integrating and addressing And then you know also you're going to that service and application layer stuff You know with the own app project that just got that just got launched with the merger of open o and e-comp You know, I think some of those business slicing Layer capabilities are getting implemented in there and we're working really hard to work on that sort of you know infrastructure Network and and back in integration piece between opnfv and own app right now All these are enablers I'm going to allow Ling Ling Lee to talk about one up. She's expert here, but The these are they are the enablers what you ask questions like Fido the open daylight openshack They're the enablers to create the network slice Yes, but they add I think lbfv is is actually acting as an integration project and it takes from upstream and do the integration testing and For registration for 5g actually own app has been targeted as address that use case But now it's you know for for release one I believe but we are actually doing the architecture design discussion hopefully in the future proof manner, so 5g will definitely be taking consideration and and I think that there is a Also special requirements from 5g scenario is that you know from the current Architecture and discussion we see that 5g actually requires two layers of service Justation in addition to service or justation so that is exactly why we're we're actually proposing that own app to be take serious consideration of having Yet another separated service or just ration up from the global like view Which is to taking consideration of the course to may 5g or just ration from the excess network Transport network to the core network Very good Right. Yes. Yeah, I'd like to continue to talk about this topic about slicing power Because I think a 5g is most driven by 3g pp Mostly and we are talking about open-source how they work with 5g. I think the Balaji has mentioned slicing is like the three layers For the service, I mean slicing and slicing supplement managers So there are two layers in the In the service layers, but if you look at the e-com I think they're just one layer about service or station, but we look at the open home. It's more like two layers. That is marching to 5g architectures. So if these two architectures are not I mean converged They are not aligned with 3g pp 5g Slicing architecture then that's something will not be I mean really aligned between the standard and open source So what do you think about? For example from Ericsson point of view, do you think the e-com just one layer of service or station? How can they marching to 5g? I mean slicing architectures and what about mainly what do you think about how the overall if not being accepted Then how they all never can support 5g Thank you And that potato do Yeah, without taking into different, you know angles here about open sources But I'm just from a purely technologist perspective, right? To be honest with you. It's not one place. I mean really if you peel into the box. There's always a you will see A box a function within the box That creates network slicing with operationally meaning operational characteristics a function that will take care of you know You know defining Based on what the service layer is telling them in task our template It is decomposed and given to the network slice orchestration level It doesn't matter what what we call the box in different areas by different companies But end of the function is the same and that will actually deliver key characteristics that is required For that piece, right? So it gives that flexibility Then the next layer is the business layer where you define the services That you could have different service models different charging models different slas So it doesn't matter those two boxes sit in one box. I don't see upon me is a technology You know, so that's what I see it Yes, I think you know the requirements. I think first of all we need to get aligned on the use cases and how we You know enable different use cases at what time frame so for now we might start with a simple use case, but I think you're right. We need to do the extractor design and have that discussion in a future proof You know like manner and looking at our 5g scenarios more complicated ones into our design from from the first day and As Balaji said, um, we we're actually having those discussions right now how to merge, you know from e-coms on current Attractor and we's open all I think open all start with Far more complicated use cases than e-com does and right now we are having that discussion and I think we are on a good way, right phil and And also we were taking absolutely consideration of 5g scenarios Because I think that is something that we need to do Right starting maybe some sometime up next year When the 5g standards are stable, I think Next half of year and and that will be really close for us and we need to Take into account that requirements and consideration right now. Yes So if I allow me continue, I think the Both of you answer very very impressive. I mean But I do see the gap. I mean because the slicing of the two layers But look at the onam today is still I mean if without include open all It still is one service ostracian layer. So if we still define the black box Say either include one layer or two layers But I still see because why they divide two layers because they're different vendors, right? So that's the requirement if we just put into one black box Include everybody I think that they're not going anywhere So that so my recommendation is for the service ostracian layer We do need two layers for the slicing. Otherwise 5g might not be supported by all Thank you. Thanks. Thank you very much. Yes Yes, no, and that's actually a really good illustration I mean so obviously a lot of folks in the room are open stackers And open stack plays a very key piece of what we're going to be doing to implement 5g through NFV and so forth But ding brings up a very good point with regard to how in in the telco world the The meshing and the alignment between The standards organizations like 3g pp and the open source projects that are implementing against those or with those as they're happening in real time is Something that I think this group of End users has to deal with more than what we've seen in the past throughout the open the open stack ecosystem And so it is it's it's a very important area on app ecomp and open o are all relatively young On app and ecomp before they merged to this new project called on app And so those are questions that are again that development community is directly grappling with right now To which open stack is a key component within that stack Other questions? Yes In that space on app is not alone. There's open button as well Absolutely So if you look at on app as a merger of the ecomp driven by 18 t open o by china mobile As my friend mentioned merging these together could be much more difficult than going another way like open button For for rest of us, right? I mean to see if these cars both compatible for network slicing for 5g or maybe another way Well, I'll I'll let others answer. I've I've got an opinion there, but please so From my perspective working with open source The most powerful thing you can do is get an industry behind a common platform This is why open stack has been so successful. Um, it got an industry behind Enabling a platform that everyone could then build upon I see own app particularly with the merger between two different orchestration projects that had been launched separately By combining those I'm confident as I always am when you put a bunch of developers in a room They come out with a really solid solution And that's what the developers are literally in the middle of right now between merging Open o and ecomp But when you get such a significant portion of the industry all rowing in the same direction with the foundation It's just helpful to get past that point and get everybody building stuff on top of it more quickly That's my perspective From my perspective, um within opina fee we tend to be fairly Broad and big tense in our sort of approach too and that you know, we enable multiple different sdn controllers in our last release We we you know did integration work around open o we also had a project around open baton integration But then it does illustrate a little bit of what you know, what phil was saying is it's just the open baton community Was a little small and they weren't able to kind of push the project So that integration work through to through to completion. So I mean, there's just a lot to be said about A critical mass of people congregating somewhere Because you can always solve software problems eventually if you have enough of the right people working on them in a healthy culture with a good You know good support So I mean I kind of feel like you know Wherever sort of the coalescence happens, you know to your point developers in a room will will figure out Figure out the problems and we we need to have sort of that that critical mass together and the other thing I always like to You know pitch since there are people here Bring some test cases too, right? You know, I mean there's a lot of good design work and stuff But you know, especially as a project that focuses so much on those end-to-end use case testing You know the more tests we have that sort of demonstrate requirements Then we can see where the gaps in the platform work where it's falling down Where perhaps different solutions work better than others You know, that's that's always a possibility. So you bring more test cases, please Thank you So I think the opnv is historically more like a manual aligned But look at the the future 5g supporting like on-app supporting them. It's different from manual If you look at that application controller, so what do you think about how how this Being different actor to be converging the future. I mean the it's just supporting found in the future. I mean how all now an opnv They are not totally I mean today's aligned architecture. What do you think about this? Yeah, so I and I basically think that there's there's development work that needs to be done You know, one of the first things we're actually setting up with them Is trying to get ci cd automated ci cd pipeline integration coming from on app Into the opnf v ci cd pipeline, which also brings in from fido and open daylight So that we can so we can just start actually laying down that full stack like on a nightly basis in doing some tests against it Yeah, so that I right now from from my perspective the the ci cd pipeline integration is actually Where for me the you know the emphasis is on because then we have the tooling and we've started down that automation path Of being able to however on app goes. We've laid the foundation for For for testing that Actually from my perspective, I didn't see you know, you know those two approaches diverse that march and you know in terms of opnv and open stack platform both, I think manual or at cnf manual Architectural or on app architecture actually treats like open stack as an external third party Component and what what happens inside is actually irrelevant to opnv integration test At least from my understanding Opnv just expands the scope to include the manual Right, so that means they include the manual, but all now it's like more like application controller centric But if so if I mean fio could be included that should be aligned But if not, then it's purely different I mean starting, you know, if you're looking at you know from the open stack point of view are those apis are actually Partial or subset of the you know the northbound apis so I didn't see any difference Very good. Any other questions? All right, well, then I'll continue on Um, so lingley all of this change uh to be able to support 5g sounds like a rather huge undertaking for the telecom service providers What are the biggest challenges and major areas of focus in china mobile for this transition to 5g? Um, yes, actually, uh, we have been working hard to enable these Network infrastructure, especially enablement technology sdnv. We think that that is actually on you know Pre-requirements for for 5g scenario and second of all as I mentioned earlier. We see that, you know, we need industry standards to be ready Um, so we're also working hard in street gpp and we were actually also Leading that discussion and looking forward to at least get, you know, those specifications Finalized I think sometime around next year and also Because 5g has been, you know Specified is the microservice like agitact application and this is also the first one is to gpp so, um, we're also working with vent Our partners to kind of like to figure out what type of microservice architecture is needed for 5g use cases Right now we have two alternatives One is very much like the nfv of the virtual machine like banner. We have different network functions that deploy it It's in different, you know, like containers, but without any like further division And into a microservice more More final granularity Or basic functional pieces But another alternative is that, you know, we're still drilling down another like level down and really divided and decoupled those sort of like Big functions into small pieces and that would require I think huge Transformation both from the infrastructure layer and also from the application vendors point of view how how they can't align And be intervable between those different tiny pieces In addition to what lingley has mentioned It's not just about the technology as well for, you know, the you know, we are talking to several operators they tell us a lot of things Related with their business processes in the people's culture and religion within the operator community Remember these are 200,000 300,000 people almost a small country And you know moving that whole force You know, you can have the best technologies in the world But people has to be retrained, you know, the process needs to be changed You know, imagine we deliver it takes three months to do l3 vpn service, for example, right? And you're expecting going to the new industry with the new business marvel The expectation is deliver the service in 15 minutes. And the worst thing is my cmo group will come and tell me, okay This product is great. Let's launch it to the market. And then well, we're not getting a good returns on this Let's close it down. So we should be able to wind down as well So that you know, so all the business process needs to be ready for that The technology will be there, but that's many operators worry about that and they're working on it to transform their force Very good. Yeah, I think network ops culture is also going to undergo a significant transformation You know the traditional mindset of you know a net network ops guy is I want to get something stable and never touch it Um, you know, don't don't sneeze, you know near it because I don't want any downtime But I mean I I really think the sort of integration of Dev you know dev ops, you know sort of style pipelines and managing your infrastructure on a much more regularly Updating basis as well as the applications on top of that infrastructure is going to be absolutely Critical for folks to actually get the the real returns on investment that they're looking for I think from NFV in general and certainly in the you know in the 5g 5g space as well Yeah, that that is something that I see as well I mean if if you remember back in the day of the central Database systems of the enterprise and how you couldn't touch those they had to be scale up mainframe type hardware To when the enterprises move to scale out linux boxes Um, you know that type of change in thought processes now going on with The operators thinking that they well, there's a big debate, you know Can you do dev ops in in carrier type networks? Or is that just too unstable and dang I I'm afraid I'm not going to let you ask a question because we're actually out of time But but anyway with that thank you very much everybody and please help me in uh thanking our panelists Thank you