 Good afternoon. Well, I'm so happy to see so many of you here. Welcome to the OpenStack roadmap panel. And today we're going to talk about what we want OpenStack to be further developed to meet telcos needs. And today we have prepared a very exciting panel for you. We have invited five esteemed panelists here to share with us their experience with running OpenStack and what their dreams and hopes for OpenStack moving forward. So first of all, I'd like to introduce all of them. My first panelist, Beth Coventry, is a secure cloud interconnect product manager from Verizon. And Beth is also an old timer at the OpenStack community. She's been a track chair for many, many years. So she probably knows a lot of you here. And second panelist, we have Tobias Ford, who is the AVP of cloud technology strategy and planning from AT&T. And AT&T is also our super user. And they have shared a lot with the OpenStack community in the past. And the third panelist, we have Mr. Ishiro Fukuta, who is the chief architect infrastructure from NTT. And NTT has been also sharing a lot of their stories with us. And the fourth panelist, we have Chris Donnelly, who is a director innovation from Cable Labs. And also Chris is also a board member from OpenNNV. So if you have any questions about OpenNNV, Chris, I'm sure he can answer all the questions about OpenNNV as well as OpenStack. And my fifth panelist is Mr. Jetsu Kan, who is the OpenStack architect as well as evangelist from SKT. OK, and I'd like to introduce myself. My name is Annie Lai. I'm with Huawei IT product line solutions. And I've been with Huawei for three to four years. I've traveled over 30 countries, visited various telcos, talking to them about OpenStack and their cult strategies. And I can assure you that at least all the first tier telcos have either deployed or looking into OpenStack. And they also have a quiet experience from the past from VMware as well as Amazon. And now they are looking at OpenStack because they think that OpenStack's interoperability is going to be a long-term platform strategy for them. And I don't want you to hear from me. I want you to hear from the users directly today. So that's why we have today's panel. And I have prepared some questions. And first of all, I'd like to have each one of my panelists talk about what their company is doing with OpenStack and their experience in using OpenStack. Would you, Beth, would you like to start? Sure. Isan? Oh, OK, great. So Verizon is definitely moving toward OpenStack. We don't have any deployments in production yet, but that's coming. And OpenStack is attractive because it's open source and it has great deal of flexibility. So with that, we can't say we're super users yet. Yes, at AT&T we've been using OpenStack in production since January of 2012. Our deployments have evolved a lot from the beginning. We're now in the process of building out the infrastructure, the sites needed to actually make NFV real. So we are in the process of deploying many sites around North America and then a bit in the global footprint that we have. Yes, we're making great headway pushing VNFs to run on OpenStack. It's very difficult to say about the entire entity group. But as you know, seeing the keynote, that we are at least like we provide public cloud services based on OpenStack. And there's a various like we use as a service infrastructure for the other entity group companies. Definitely private cloud is the foundation for R&D. So lots of component or efforts we are consuming from the OpenStack. And Cable Labs is an R&D consortium charted with bringing innovation to the cable industry. So we are using OpenStack as part of our research and development program around network virtualization. Virtualization network evolution as we call it internally. And so it's a big part of OP NFV and we're using it for our NFV infrastructure to bring up our virtual network functions. SKT, we used to have public cloud and private cloud based on the vendor solutions. And we decide to go for the OpenStack this year. So we are currently deploying public and private cloud based on the OpenStack. We'll open the beta service at the end of this year. So it's like a new start from SKT using OpenStack. And in addition, the SKT think the 5-inch network needs very programmable and virtualized and very flexible infrastructure. So we try to build a software-defined technology and software-defined data center operational platform. And we are integrating OpenStack Keystone as a unified authentication method and looking into some neutron to be used in our operational platform, so. OK, great. Well, I'm so happy to see that there's a lot of OpenStack experiences and practices here on this panel. And so before we get into the roadmap, we're kind of wanting to know that what you have learned from your experience with OpenStack, the good in the bed and the hopeful. And since I think AT&T is probably the longest, has the longest history on this panel with OpenStack. So Mr. Tobias, would you like to share first? And then I'd like to go through the whole panel and hear about the rest of the panel's view. Sure. So I mean, I think the biggest learnings over the last three or four years of working on OpenStack is definitely having to focus on the team that you use to architect, to implement and to operate, to implement DevOps with. Getting the right folks involved, having the right partners is really an essential ingredient. Because, and I was talking about this yesterday, OpenStack is such that it's very easy to get overrun. You start, you set up a deployment, you get two or three setups in place, and then you have so many applications and V&Fs running on it, and so much demand that you really have to be ready to scale. And so that's, I think, probably the most pertinent lesson in over the time. OK. And is Cheryl's done? Yeah. Oh, yeah, go this way. So our experience in the OpenStack is how OpenStack community is an evolving project, right? So each time we see lots of proposal and keep feature upgrading. So our development starts from 2011 years. That's still, if you look at the neutron, it's evolving. So each time we have to be careful to which feature to be used and how far ahead we can start working with the community to talk about the feature of what we as a service provider need and a co-developed feature with the community to do it right on time. Sometimes it doesn't meet, but we had a good idea about dealing with those open source community in entity. OK. And Chris? What we've been finding at Cable Labs is how OpenStack will help us accelerate our internal development. We're using CI-CD system that's tied in. So we're using it primarily with OPNFV. And so we're tied in with the Linux Foundation integration tools and then bringing in OpenStack through that door. And we're now using virtual teams going across the company, across different departments within R&D, and then also bringing in RIT support staff. And we're able to develop on a six-week cycle. And for us, that's something we could never do in the past. Previous projects, I think the fastest we were able to do was about four months. And beyond that, we're finding that we can reuse the same code for multiple different projects. And so every time we start out a new effort, it builds on all the lessons that we learned in the past. And we can move it that much further, that much faster. Anything negative? I mean, it looks like you're very happy. OpenStack uses it for your internal IT. But is there any lesson or kind of culture that people should be aware of? It doesn't solve the full problem for NFV at this point. We're still missing some pieces, management and orchestration from a service perspective, Layer 3 networking concepts, and more focus on the access network rather than in the data center. So there's still some work to do. But we're happy with the direction things are going. Well, he kept knocking his head. It means you agree with it, right? You agree. So would you like to share your experience with the good and the bad and the hopeful? I think the deploying OpenStack, I'm not saying that's the easy thing, but most difficult thing is operating the OpenStack. So when I talk to the node guys, the network operation center guys, they don't understand cloud concept of the architecture. So I have to appreciate them. Starboard guy and network guy has to put together and see this OpenStack as one platform. And there's lots of operational requirements from the traditional telco environment, which is not right for our cloud environment. So talking to the operation team and make them understand and prepare some features ready for the operation, that was most difficult thing that we are going through right now. And at SKT as a telco, we are not used to do those Azure development and CI CD stuff. But when we are adapting OpenStack, automatically we are accepting those DevOps concepts. And development process has to be in place in the SKT. So we are also working on changing that culture inside company to be more DevOps-centric approach. And that? So I'm last, but we're probably the least far along. In our journey into OpenStack. And I would say probably the biggest hurdle that we're running into is the cultural, the need to change how you view tools. And you mentioned agile. You can say you're agile, but that means something different when you're actually doing it. And I think that's been a real eye opener for a lot of people in our organization. And of course, it's a big, we're telco, which means we're a big organization. We're sprawling. And we have to deal with a bunch of different groups and how they work together. So it's been interesting. So it looks like on this panel, we have used OpenStack for private cloud, for your internal IT development, for public cloud, and also for NFV. So in your opinion, which kind of cloud deployment OpenStack is most ready for? And which one is least ready for? Who would like to take a stab at that? I think NFV is pretty clearly the least ready. Sorry, layer three, that's something we really need. MPLS, we really need that. And it's not going to get very far if you don't have those components. Probably it's a toss-up between, well, I'd say public, private cloud is probably the most far along because it's the one that's got the least sensitivity to bugs. You can kind of live with them more than, let's say, public cloud or full production types of things. So certainly for us, we're still definitely in test dev mode. So it's definitely ready for that. But sort of the next level, you know, those tools have to be there, they have to be robust. Those five nines really mean something to us. Right. And Toby? So I totally agree with what Beth was saying. But in addition to that, I think, in a little bit of a different dimension, when the app is ready, whether or not it's an app or a traditional IT app for public or private consumption or VNF, when the app is built to cloud-native methods, then it can work. And it can work on fairly old setups of OpenStack. We have actually still versions that are quite old. I won't say how old, but very old. And then there are apps that run there that have been up and running 100% since that time. And so those apps are typically ones that scale out that have elements or components that are in multiple facilities across multiple zones of OpenStack that have distributed databases, distributed NoSQL databases. Things that are more toward the microservices model, more toward a completely automated, have full change management around them. Those types of apps work very well in OpenStack. As I said, they could run very well over time. But to a best point, there's a few things we need to work on to make it fully ready for VNFs. So I agree from a neutral perspective, OpenStack networking perspective, and if the capability is not much or yet. But speaking from the, not talking about OpenStack as a services, currently I'm doing, we are developing some software-defined services using OpenStack components inside the product itself. So in that perspective, to deliver the end-to-end software-defined WANN services, we can use NOVA, we can use WANN's Keystone and Heat if the neutral is not ready for the NFV. So there's a certain module that we can consume and make use of it and develop the services on top. So I think OpenStack community or component is extremely providing us the modules to develop the services, even including NFVs. So if the community is missing some NFV features, but we can work with the vendors to have who can deliver us the SDN NFV service-changing functions at the product level. So there's lots of integration patterns to deliver the product to the market for now. Okay, Chris. And I agree with what everybody else has been saying that NFVs is still very immature. But that said, the community is working on it and it's been great through OpenNFV to be able to work with the OpenStack community and make some of these improvements. And then I would say coalesce around the need for some of these changes in future releases. We had some very good meetings this week on Paths Forward for fixing some of the Layer 3 problems or the management and orchestration problems that currently exist. So I think that we can get there. It's just going to take some time. So we're very lucky to have you here because you're not just because you understand OpenStack, you are also a board member at the OpenNFV. And there's a lot of talks about, you know, there might be some overlaps or there might be some kind of conflict between the two organizations. Can you give us any advice on how OpenStack community can work with OpenNFV community so we can all move towards the same direction and, you know, make progress a lot faster? Sure. OpenNFV is a midstream community meaning that we don't own the code ourselves that we're working with upstream communities where they exist. And we work very actively with OpenStack in a number of different projects here and then Open Daylight and Onos and a number of other components within the ecosystem. So we don't see ourselves as competing, if you will, but collaborating upstream doing the integration work between all the different components, finding out where there's friction and then making proposed solutions within the upstream projects to make that go easier. Excellent. Jesse? Yeah, I agree with you already. As your question, private cloud, so we can have some level of control through what kind of apps and processes and work load we can host. So for the OpenStack to build private cloud that not its task but relatively say that's more suitable for the public cloud, it's not the question of the technology, it's the question of the business. So the business guy is asking what will be a differentiate service feature we can provide to compete the other public cloud in the market. So mainly that tends to be focused on the network services and the security around the OpenStack. And I'm really happy to see there is more network services and projects in the OpenStack community, but that's still just a growing project. So it's kind of very difficult to apply those like the VPN and firewall and service into production for the public cloud. So that area, I might be very interesting to make it the faster and stable and we are willing to contribute if it's possible. Okay, excellent. And I know that moving forward the foundation, OpenStack foundation is also thinking about building up this application ecosystem. And for telcos coming up with the innovative telco digital services is very important. And telcos is also trying to visualize network functions and services. And OpenStack has incorporated Kubernetes and containers and all these things to help developers develop faster and deploy faster. Do you think that's enough? Or do you think your guys are still going to Amazon? Who would like to take a step at that first? Sure, great. Thank you. I think, yeah, so OpenStack, so as a service provider, so we need to provide lifecycle and like a day-to-day operation. So like having more monitoring visibility and supporting orchestration function is critical to make those OpenStack or any SDN, NFV service component to orchestrate to build the end-to-end services. So we'd like to see some more, having more focusing on upside to evolve in this community. So that will give us more like a better, like a consumable. More tools as well. Yes, that's definitely help. Tobias? Yeah, so I'd add to that two parts. So you're talking about containers as one, I think the work on Magnum and Kola and the other projects that have brought containers to OpenStack are good examples of how the innovation in the model of OpenStack works. It actually, it does move forward. It doesn't stay limited to just a small number of things or functions and it's able to evolve. So I think that's actually quite exciting area. And then as I was saying, we've been promotive of the cloud-native move and that's an essential ingredient there. Not just to be a bit more efficient than virtual machines. And certainly there's a lot of work to make containers secure and such but I think it's great the movement toward something simpler, something more transparent and easy to sort of manage. That'll help us in many, many ways. As the environments get more complex we have to fight the urgency, the urge to make it more and more complex. We have to clean it up and refactor. So that's been a good thing. On your other point about the Amazon piece I mean it's always a constant threat. I mean given where they just recently announced the $7.3 billion US dollar business, that's pretty significant. The scope at which Amazon can solve problems is growing. And I mean so that's something we have to be mindful of and really be competitive with. If we're not able to be competitive then people move off. So it's essential for us to keep that in mind. And then also when it comes to telcos though I think we'd all agree that still there's a lot to be able to do what we do when it comes to access and connectivity and often that means having facilities very close to the users. So in North America between us I think we probably have a quarter of a million buildings. And so that's hard for Amazon to compete with. I have a couple of things. First of all I think organizational readiness is pretty key. I mean again we're relatively young in the open stack journey so I see it very painfully at this point and I don't think open stack is as friendly toward companies that are just starting in that journey. And I've been involved in the open stack community since pretty close to the beginning so and it's come a long way. Good news come a long way. In terms of Amazon I'm not I actually work very closely with Amazon I'm not afraid of them at all. So many people have heard me say this. Amazon's target user is three guys in a garage and they have an amazing lack of understanding of anything beyond the data center. Now they're great inside the data center but as far as they're concerned it gets to that router and beyond that it's just magic. Well we know on this panel it ain't magic. There's a lot of hype about Amazon. I've heard that too. Who else want to address that? We're I think containers are very important we're using them a little bit differently I think in our virtual CPE development part of the architecture is to allow some of the virtual network functions to live at the edge we're using containers to push VNFs down to the edge where we're implementing them on top of Raspberry Pis which are very lightweight CPUs and I think that's a key differentiator from an Amazon that really is based on data center operations so we've had to do a bit of training to get our container strategy to work for us with the current tools but as OpenStack further develops the container projects I think that's going to help a lot for distributing them throughout the architecture. I want to share your thoughts. Okay imagine for the SKT as a technical we have benefit in understanding network and providing network in the edge side but since Korea is a small country just one or two data centers can cover the entire country with that latency also the AWS is a global service and as a technical we tend to be the local service provider but the user using the cloud service usually requires to be a cloud service and in Korea we are trying to figure out what will be our strength and benefit can be provided to the user as a public service provider so that's our big challenge and question right now we have to serve for the internal I agree the container so amazingly in SKT organization we start talking about MSA because we just start talking about the CICD then we talk about MSA as well so it's a very big jump because it's all because of the container and we can use the open stack because of the project related to the container to provide infrastructure to develop our culture toward MSA our applications architecture services so that's really good and beneficial for us to have this open stack not as just the infrastructure platform sometimes it's like the integration engine as Mark said to have lots of different of technology but there's central point we can have to talk with lots of different vendors and that's very beneficial and helpful for us okay great just want to let you know every time I turn around there's more people so just want to let you know the audience you all have a chance to ask questions or give us your thoughts about how open stack can help telco and how open stack should be evolved in order to help telco compete and do better so coming back so I'll come back to you to start thinking about those questions and later we're going to need you to line up behind the microphone so you can ask questions there so coming back I'm so glad that Beth said you're afraid of Amazon that is the right attitude and Chris you talk about differentiator so I know that telcos use open stack for private cloud for efficiency so tools are very important to help you develop faster and with more agility and then for public cloud some of the telcos are in the public cloud business then that's where they compete against Amazon or some regions Amazon are not there or because of data sovereignty issues they can't even get there and at V I don't even know if Amazon plays there yeah they'll never well so from your perspective how is open stack able to help you differentiate and compete against Amazon and if we are not ready open stack is not ready what else needs to be done or needs to be further developed in order for you guys to compete against Amazon or Google or Azure I'm sorry Google, Amazon and Azure are not telcos yeah they are not and they're not getting in the telco business anytime soon right alright I think on your point about what does open stack need to do to be more competitive I mean clearly to enable you to be more competitive I think the BT guys were very clear about this recently there are six or seven things that they feel like are essential ingredients and I totally agree with them the one thing I'll focus on is upgradability for not just taking something and doing software upgrades but actually making a commitment to CICD and helping everyone and make CICD real I think that's probably for me the most important one because then like Andrew was saying we can't add more functions we have to wait if we have to wait for whole new versions of open stack we have to go that way and if we don't have the right sort of CICD test framework around it then I think that's going to be problematic so in the open stack world we spend a lot of time on Tempest and then in the opnfv world we spend a lot of time building the right test harness the test framework around the specifications that have been provided and I think that's that will help to get us going in the right direction in this way so I think more universally so I think customers journey to the IT of course like it always packet comes into our network first before it reaches to that AWS Amazon or any other cloud so if you look at the enterprise users so there's lots of people in the market that it is not still like a not starting even start adopting the cloud services so we can in this space business we can help our enterprise customer to onboard to the cloud from starting from the network so definitely integrating NFP or like a good experience using open stack and virtualization technology to change the experience of the IT to the end user will be the differentiator or we can wrap it around so I think the journey should be considered from the end to end experience Chris? I agree with the chair the network and the proximity to the customer are key differentiators for telcos as compared to the over the top cloud providers and I think for open stack the key to enabling that opportunity is moving outside the data center in terms of the network support really think through layer 3 think through VPN connections and it's not just the XLAN we need to go further Yesu? For the customer it doesn't matter if it's an open stack based cloud it's much more like a service feature they really want so I'm totally agree with Tobias we need if we want to develop a new feature based on the open stack we have to be able to do that as quickly as possible in a very stable way so I think Tobias comment totally agree with that Okay so we have about 8 minutes I like to open up the floor and see if there's anybody who has any question for our panel or if you have any kind of insight you would like to share the brave one I like that so you would like to share with us that you think how open stack should be further developed in order to meet telco needs I'm going my own way Okay great and tell us your name and the company you work for first My name is Greg Stigler and I work for AT&T I'm a peer of Tobias this has not been choreographed he may be horrified at this point so here we go the British telecom article hit home with me agree with the six points like Tobias does as well then I looked up stackolytics I don't see British telecom playing so it seems like a manifesto to me versus a community participation involvement now I also want to admit something else AT&T has had a lull in our contributions as we have developed our capabilities and I will announce today that we are going to exponentially increase our contributions to open stack beginning now what I want to ask the panel is I've heard several times up here the community needs the community needs to do this the community needs to do this what about the telcos what's our responsibility to the community thank you I will take a seat and listen for your answer and I like to comment if I can I'm a moderator I'm not supposed to comment because I work with a lot of telcos I do totally agree with you telcos are using open stack but they are not necessarily part of this community because they figure they have vendors who represent them and I think we need to change that for me I make it personal goal that I like to make my telco customers super users because they speak for themselves instead of having me talk about their use cases and we need to incorporate them because they are the ones who really know what they are looking for and so we need them if not to be contributed but at least to be active users and then can give us feedback so I hope that anybody who is a vendor here would do the same next so Gavin Pratt from HP and so I heard a number of you talk about containers oh sorry I'm so sorry actually I heard Toby I would actually urge itsro to talk about his contributions because it's pretty exciting what he's been working on generally the network part of open stack we've said this many times is not sufficient to what we're looking for and so recently as I think Chris was mentioning we're really trying to work together better now we had a great meeting earlier this week with all the tocos sat down and talked through on a particular issue I think we're seeing a lot more momentum now some of it may be just because the hurricane Margaret is involved in this project but but I do want to hear itsro's stuff though it's very good we had a very good discussion this early in week about what's lacking in Neutron, what we should be done so I think more and more so we need to converge to discuss about the model and the future what we can develop for instance we will open source and do the now step like a source development framework today after like 430 session I will give the model driven and database kind of software development framework to model the network services on top to contribution so we are open up to discuss with you guys to make service providers into the software development okay good or I can add a little more so SKT is kind of new in the open stack area I I've been in open stack for like from the beginning but yes SKT is new but like a financial and technical it's too most closed industry so far but with open stack it's changing so SKT is like opening our SSN contribution for example SKT is working with onus and we decided to contribute our code base and contribute developers and code to the onus community so that they can have those integration between onus and open stack it will be same in here open stack so we are now just starting but we are willing to put our developers and really give a code contribution and it's necessary for us I mean we have to contribute for us otherwise we will have a silo version of onus which only us to using open stack so it's changing let me say so I think it's a cultural shift thing we have not contributed much to over stack if anything other than me I think I'm kind of the lonely evangelist and it's really down to the company doesn't really understand the whole concept it's a new it's a paradigm shift and it will get there we will be contributing at some point but we have to get there and I'm sure AT&T took a while to get there too and we are in the process we are still in the process of ramping up our resources Cable Labs is a small company we have 200 people total 100 in r&d and so we have to be very focused we are going in mostly through opnfv to work with open stack, with open daylight with the rest of the community and currently we are leading one project in opnfv and two in open daylight plus contributing additional test resources and then participating as we can directly with open stack at summits like these okay so I think I only have two more minutes left so we can only take one question and I just want to before we get to that I just want to put the word out there we are going to have another panel who is going to be hosted by Beth she is going to put out the vendors talking about nfv stick around it's going to be more exciting than this I'm sure so Gavin Pratt from a vendor HP no question about containers I heard a number of you mention them especially when I talk to the telcos I hear a big mix of scenarios under which they want to use containers and it would be useful for vendors like us if I could just get a quick vote down the panel which of you want to do just containers versus some kind of management technology like the kubernetes or mesos on top versus ones that want to do a full pass like a cloud foundry managing those containers and then if we had time I'm curious when I hear that more basic container scenario a lot of telcos they start going down that path and they decide they want bare metal instead and so if you had time to weigh in almost on like a yes-no basis for each of you it's still up in the air I think we've tried them all yeah we're just at the beginning really figuring things out that way and we're pushing on the VNF vendors so you should see I did a demo yesterday with Ericsson and AppSera folks on one component like the one piece of the IMS system but there's a long way to go before we could get a production grade VNF with containers well I'm sorry before you so there's I mean when you say this up in the air are you literally considering everything from just a Docker container all the way up to a cloud foundry orchestrating all of it or anything out of bounds or beyond the basic part of it there is more you need more of the orchestration piece to make it work right and the Kubernetes Maze's work is very exciting very promotive of the stuff that we've done so far but then the next layer up and I think that you'll see this I don't know what it is but in the last two three weeks we've heard a lot of discussion about service orchestration and a lot about MANO and the part of Etsy that relates to management above what OpenStack or VIM would do and that I think you're going to see a lot more talk a lot more collaboration and cooperation about how we do MANO so container is very interesting but it depends on like where we fit like it's just using Docker container or like having Kubernetes Maze to orchestrate like manage those containers so it depends on the where the use cases are and we always like if we can choose the right and try to make it simple as much as possible so right we don't see a pure play for containers we're looking at VMs containers and bare metal in the data center and then either containers or bare metal at the edge and like the other panelists we need a common management platform for all of those different scenarios so it needs a significant flexibility as we go forward for the for the developers internally we are starting providing PEST because they don't need to aware of the containers but this app isolation so we use PEST and we are testing out and as an infrastructure layer we are going to focus on the container and orchestration layer altogether with that we have reached the end of our session and please put our hands together and give our panel a great applause stick around for NFV Bake Off thank you so much