 Hello Should we get started for the next panel? Okay So I think we run a few minutes late. It's pretty interesting exciting talk so this panel is a It's called the experience with open stack in telco infrastructure transformation Welcome. My name is Hain Wang. I'm the CTO for cloud computing at a Huawei I've been involved in the virtualization and the cloud Over ten years in my early days from since we are more So we all know that it's pretty clear the IT is going through transformation move to the cloud and Cloud I'm hopefully will be cloud stand up by open stack as a primary platform for the future infrastructure and same time the telco is going through their transformations and and We see them like in every initiatives and cloud service It's all Exciting. It's my distinct honor to hold hosting this powerful panel with the guys who are very much in front of driving this transformations and We have opportunity to hear what the telcos unique challenges and efforts From this practitioners so in the panel what I'm going to do is that I'm going to Ask each of this panelist to introduce themselves and what they have done in terms of open stack Then we'll ask a few prepared questions We leave a few minutes in the end for audience to participate So we start a lot of delay. Let's get it started so Hi, thank you guys for being here and Another Boris started and then you can pass the microphone. Yep. Thank you. Hi, so my name is Boris Rensky I am a co-founder and chief marketing officer at Morantis. We are the largest pure play open stack vendor And my role here. I'm unlike some of the other Members of a panel here actually represent the vendor side. We work with a lot of carriers helping them Adopt open stack for the NFE use case So I'll probably serve more as kind of a co-moderator with highing then kind of representative of the kind of telco viewpoints But we'll add some flavor of my own as well Thanks Boris Yeah, I'm Toby Ford. I work for AT&T. I'm responsible for the cloud architecture for both our internal and our external work and then I've been with AT&T over the last eight years. I was actually CTO of the company USI before that and then Lately we've been working a lot on the idea of NFV and SDN applying it to the telcos So my name is Alexis Sassets. I'm chief cloud architect at Vodafone and also Product owner of our private cloud platform So I work on the application side of the business always try to bring Sort of application view to what we do in the cloud. So looking forward to explain more of that I'm Fred Oliver at Verizon. I've worked in the corporate technology group Worked in various cloud areas. I was acquired into Verizon from a small company called clouds, which I've been working in cloud area for Probably eight or nine years different forms And I look forward to explain some of the work was going on at Verizon Okay, thank you. And let me start it with the first question. So let's first talk about a cloud So I want you from your experience Why what force driving you to go to the cloud and why open stack? What's the experience? What's Challenges for you specifically and using open stack. Maybe you can start it from Fred so that again, what's driven us in this direction is an attempt to change from the single box the deployments that are Monolithic deployments that are happening in our industry all of our vendors tend to give us a single environment That is distinct for every single application. And so our operations group is huge we have a No common environment that our operations team can learn and so we're trying to move that Goalpost a little closer to a common environment open stack looks like a very good Approach to solving that it's not completely there, but it's a pretty good start All right On our side really what's what's been driving us towards cloud is I always reason that to agility So it's really about making the business a lot more agile and an example. I often give is that not every product is successful And we also Well, we often saw the very strong investment into it to support an application That will go to waste if the product is not as successful as we wanted So this agility brings two things. It's really quick reaction to activities in the markets And the quick decision-making in terms of you know Do we keep going with that application and change it quickly or do we just call it quits and do something else? So that's really the power of cloud what we like with open stack is really It's becoming the de facto API And that's really what's driving us towards open stack. It's we hope it's going to be a standardization across all areas of Tercos Yeah, so I think the thing driving AT&T is Added to what Fred and others have said Alex if it's already said is Is a certainly full asset utilization and consolidation, you know typical path for virtualization also Giving the users real-time experience right so your our customers are wanting to get access to the functionality We can provide right now. They don't want to wait and that's a Change for the telcos is there's a lot of operational and user experience that's been quite slow So switching to real-time is important that way I think in terms of open stack We've been using it for quite a while for about four years for a lot of internal applications But you know it's been as described It's good for standardization and I think also as you see lately the broad set of functionality that's showing up in all the Adjacent projects the projects that are growing out of it some of them are very targeted toward our needs so things like Congress for policy and some of the the messaging and workflow tools those those are really target systems that we have to build so that's That ecosystem that federated and sort of integrated ecosystem is quite quite helpful and for us Some of the challenges we've had obviously with with the open stack have been the life cycle of working with upgrades and migrations and just keeping up with With all of the great innovation that's happened but trying to be on the more recent versions and that's Trying to pivot toward more of a CI CD model has been really hard for us within our ops team. So and then obviously Performance through the network is a big area for net for telcos trying to make an x86 perform at the kind of packets per second that we're looking for for our Applications it's gonna be tricky. So that's those are some areas Well, like I said, we are not the telco we are a vendor so I can add some things from kind of vendor perspective So there's there's a variety of use cases for open stack in Carrier and telco space and I think that what most of the Of the colleagues to my left have been talking about agility Combining different silos into a uniform environment. They to a large extent pertain to the Kind of a via internal private cloud use case but the use case that We've tackled quite a bit is actually using open stack as a virtual infrastructure manager for NFV and I think that is Kind of a probably a more interesting and Unique to carriers use case for open stack than using it in the same fashion that enterprises uses it for enhanced agility So in in moving the conversation that direction. I was wondering how many people in the audience are familiar with Just kind of you know NFV movement in in carrier telco space by show of hands if you know what it is Okay, so good good good amount of people so the the one comment I want to say and I think that probably Toby and some of the fellow panelists can add a lot because they are kind of you know the driving force behind making open stack via via the de facto kind of Vim fabric for I think that what makes open stack unique is the fact that it's has emerged as On one hand open on the other hand pervasive API standard for managing virtualized infrastructure and Historically based on kind of the customers that we've worked with The concept of a virtual infrastructure manager is not something new they existed for a while and They've existed and in various forms and various implementations, but now as via telco space is really moving Proactively moving away from buying expensive boxes and adopting commodity gear and Running Vim's on top of of commodity gear We at the same time open stack has kind of gained its prominence and this two trends have converged and open stack has become kind of this unique fabric that that can be used for managing various virtualized network functions and Right now there is a trend that all of the kind of you know Existing Vim's that that have been around are being you know largely replaced by by open stack and we work with some of the Large telcos as well as large offenders that have their own kind of flavors of virtual infrastructure manager that they have evolved over the years that are now embracing open stack as kind of the de facto standard for that and Are also augmenting it and contributing a lot back to the open stack community to make it even a better fabric for that Okay, that's great. I think it's a good segue for the next question Thank you for an open stack and I think this is a very hard especially in this conference and I think a last Last conference and Toby have a keynote about this. So I think a Year ago I have I have one it was initially sort of Both are in the dressing the same problem, but a sort of in a very different directions One is a top-down one's a bottom-up Open stack side is a king putting new futures and telcos I did putting more wishes to them and as we go I think a lots of practice practitioner is involved a lots of project and how in the case we work with many of those wonders to on the POC's and on the same time to Open stack as a whole also put this in high priority and we see some special We have we keep page about all NFV. We have a special group of neutron for NFV We have lots of functions in there from NFV. So what I want to ask The panelist is that so maybe you can give your definition of NFV. What do you think? the challenges choose open stack as implementation platform and and how you think this tool can be converged to Since there's lots of people here some from IT from hotelco will be good to And a boss and I'm on one the side that the other three is a user cell signal I want to get this conversation goes see what's your you guys assault So I'll probably hand off the Toby because Toby is a fellow Open stack board member that's been very vocal in driving open stack in the direction of specifically making it usable for the NFV use case and probably knows more about that than most people around Sure. So just some of the things with NFV You know clearly there are systems that were were once in vertically integrated hardware bringing them out into Virtualized machine in VMs or in containers. That's a challenge. So that's the first step is is taking The software out and making it work on on a Linux box or you know, that hasn't always worked So that's sort of the first step and it's very similar to enterprise space We're taking it the legacy back-end systems and and converting them over to more of a web app database Model and more of a distributed database no sequel making it go from more of a pets to a cattle model We're having to do the same thing with the NFV kind of or VNF's the virtual network functions Convince the vendors to to think more in the scale-out model and move more toward what I think of as midget cattle Because when you're when you always thought of vertically integrated boxes you just make them bigger I have had more processors more memory more throughput through through the boxes and then if if you think of it more as midget cattle then you know You can build out look much like a Google or the big web companies have very scalable and resilient environments with a much cheaper approach and much more kind of You know swarmish approach to the problems of getting the the vendors around the the VNF's to Switch to this is been key and it's you know We're attacking it both ways going at the incumbent vendors to try to get them to change But also promoting the startups that are targeting this the spot and you know things like it's actually been around quite a while If you look at what let's say Viata has been able to do with routers For see CEs and PEs within an MPLS scheme. They've already had good penetration there for especially globally If you needed to have a cheap CE or customer edge product, they were already there same goes for load balancing and firewalls They've already kind of made the switch to being virtual Now it's a question of the the larger mobility kind of system switching so Um, so I perhaps have a little bit of a different view on an FV Then most of the other tech or people as I come from the application side of business What I find really interesting with NFV is the the dynamic aspect of workloads that we are seeing happening as Says more and more services are moving towards IP We're seeing more and more bandwidth requirements latency requirements with the evolution of the network And and I always ask myself the question. Well, how is the application of tomorrow going to look and How are we going to get this very Elastic and instant access to resources all across the the Terco space so NFV really answers that question in combination with OpenStack because We could potentially move an application from a data center To a core area and even to the edge when when it needs to happen so that's quite interesting the Challenge that I perceive in terms of OpenStack, which perhaps is not directly an OpenStack challenge But is the ability to run This kind of new workloads without affecting the the critical workloads that we would have so Emergency services called these kind of things you want them to work even if you have other workloads running on the same platform So that that's where I feel the the architecture work around OpenStack is is really going to pay off and and give answers Which actually will also benefit IT. So quite a good space. I don't have too much more to add to I think we have the the problem that we've seen and kind of in our vendors is they exactly what Toby said that there is a lot of kind of extraction of what they build in the Hardware environments today and trying to run that just in a virtualized way that doesn't allow for the Dynamic scalability or reconfigurability that you really need in this environment So I think that'll be a challenge for our vendors and I think enabling or promoting some of the startups is probably a way to move that along and Get that in a more consumable fashion for us as a consumer from a kind of Ability of OpenStack, I think the a lot of the capabilities are There are almost there and existing open-stack environments NFV is getting pushed in Pretty well, so I think there's lots of capabilities the notion that being able to describe an application now as a through an NFV NFVD or VNFD is a Useful approach that you can abstract the needs and resources into a consumable fashion for an infrastructure and deploy that in a common way will be a Enabler for our environment Okay, and do you guys have anything to add I think I think the The OpenStack site I can see two things that are interesting First the OpenStack itself is sort of a moving target with any more new features Because we're driven by engineers, so they always want to add a new things more fancy things keep changing. That's good Innovation, but from consumer side you consume this technology Do you wish to be? In the form that as some parties are stable so you can keep going or or you want just open-ended Just keep going so that's saying we have lots of customers asking saying Can we can you say it's something it's gonna be production ready? I mean what he really means that maybe two years you don't upgrade this stuff or can keep using it So this is the one thing another is that as a whole to address NFV issues Right now OpenStack is more modular driven. There's no Coordinated effort and then we have two bottom members here, so I'm trying to use this opportunity to say There's another organization called OPNFV. That's all more practical. So from Is anything you guys wish that it'd be more systematic instead of? We just let it happen So I'm just through the question there and you guys can pick it up Well, that's my perspective. I think a lot of things we need is a lot more of the integration between Kid believes the service assurance getting monitoring capabilities that can feed back into the control capabilities and of getting the right models and behavior out of you know, salameter or other approaches to that feeding back into orchestration framework that we can both Understand what's going on in the environment and use that to Feedback into how do we run it Perhaps one point from my side, which I addressed in Hong Kong together with a highing as well is really How are the APIs moving forward and ensuring that? Any changes to the API? Will not break any software that we run on top. So it's pretty basic statement, but I think it's always good to remind ourselves that There go applications may not move as fast as as IT and so they have a heavy reliance on whatever they use and Yeah, the thing I'd add to that is really supportive of the testing work that we do with Tempest and building automated test suites for each of the the different projects and Trying to come up with like what Boris and Miranda have done with the vendor compatibility. That's very important Trying to continually improve the the way we we test refactoring the kind of core and getting at the issue we're just describing of of always extending iteratively functionality you have to kind of refactor and come back to okay is the base part solid and Does it continually meet the expectations that we set and so that's that's what I think is great about open stack Is it provides an example for other open source projects of how you can build out a very comprehensive test suite to? Validate improvements and such so I think that's that's an example of helping some of that problem Yeah, thank you. So I think that so the question if I'm understanding correctly Open stack is fluid. It's moving fast and things change. How do you? kind of cohabitate this fast change with the SLAs and predictability that the consumers of open stack in telcospace need so if that's the question my my answer is as follows I think that It's important to understand that open stack is open source community and as such Everything that's happening in open stack upstream is really kind of a development sandbox Environment, this is not exclusive to open stack. This is permanent. It's typical of any open source project So if you want to take The upstream open stack And make it usable you have to spend some effort to make it work And there is a couple of ways that you can do it you can either work with some vendor that ships a downstream distribution either specifically Hardened for the telco needs or the enterprise needs or whatever you end user need is or you basically have to roll your own distro And the common way to do it is through, you know, the very sexy notion of you know Doing CI CD and running the trunk So somewhat first of all like, you know, nobody nobody ever runs really the trunk What people do do is they absorb some of the patches that emerged upstream and then they have Their own sophisticated system for testing validating fixing and then merging it into the production environments Which is effectively what you know vendors like morantis or red hat or Canonical do when when they roll their downstream distribution of open stack And I believe that the choice between either going with a vendor or doing it yourself is Driven by whether or not, you know, it's very important for you to stay completely independent or You want to kind of, you know be quicker out of the gate, but you know, I think that this this lack of predictability in the upstream environment and The fast moving pace of development is not something that is ever going to change With open stack or with any open-source software and it's important to understand that great, so Let me go to the last prepared question. I think open stack is different than any other things at least Huawei perspective as a close-source company is this open development model and it really the users Developers all in the one community to push so if for us is a pretty Dramatic for us because we always know customer order they want and then we do the future and deliver I hope they can use a couple years and then we do but now everything Is transparent that's good and sometimes you don't have a control what's going on So I want to hear a special from the operator perspective in this open development model What why is a thing how important for you as organization to have solved these technologies? What's your thought and if anything you wish to change or anything you Change yourself then you'll be shared with the rest of audience will be interesting So I want to hear from you guys open development model. How important is is this going to be future? That's the way it is It's going to be For time being to break through the new technology after stable work back to the old time style And the kind of to piggyback on that question that kind of I want to add a related question as a The question is about the degree of penetration of our commodity gear in telco space today So we talk about NFV and open software orchestrating on top of commodity gear Is this the far out kind of in the future thing that maybe will happen or is there a significant chunk of infrastructure? In telco space already today being run on commodity gear with fabrics like open stack being used as virtual infrastructure manager Sure. All right. So two questions Try to remember both So the first one about the open source and development model I think it's essential for for right now. I mean if you look at the trend over the last 20 30 years Whatever function it was. I think my in my experience compilers used to cost $20,000 Buying a compiler today. That doesn't happen It's done with open source and it evolves very in a very interesting ways like LVM and such right you do that with OSes with databases with web servers same things happened It's essentially when it's non differentiated function and everybody knows how to make it and everybody wants the same thing Then open source has proven to be a very good model for working that way I think it actually represents a better way to create standardization Than standards committees and this is one of the real reasons why we've been so supportive of opium FV Is we were trying to actually create something that's a merger of the good parts of let's say Etsy and The good parts of open stack Have developers working on practical Deployable thing Have it be well tested and integrated in something that we can deploy and work on actual code And iterate on it instead of waiting for committees to decide and and voting on it It's more like a practical sort of feedback loop. So that that we just I think Also, I'm hopeful that we get all the other telcos to work with us this way the telco space is enormous worldwide the revenues of telcos that are in the trillions of dollars if we could just apply our wasted bodies In that space to this this problem We'd actually solve quite a lot of problems and really we have to accept That there isn't a lot more differentiated function in this realm It's just not the case when there's 100 100 telcos wanting the same functions It's not that way So open is the right way to go and we're trying to drive that way and try to be as independent as possible as well You know, we've spent so much time being dependent on one vendor or two vendors And then that's led us only down into a dead end. We want it to be more independent and open this way Okay, so that's question one the second part of The second part of it is about the hardware commodity hardware. I mean not a lot of what we do today is on commodity hardware It's still even when we use x86 hardware. It's it's brand name. It's Uh done with a four hour SLAs we pay a lot for maintenance It's still not really the benefit that we see in the future with commodity hardware That's coming and that's clearly where we want to drive But we're also like fred was saying earlier We're not going to get the economies of scale Going the way we've been doing it before where each service we have is a siloed infrastructure We have to pull it together and start buying in chunks of not a thousand x86 servers But tens of fifty to a hundred thousand a quarter, uh, like some of our competitors that way, so All right, we have very good answers. I can't add much more. So, um, really I think I reiterate open source Is a great way to drive standardization And I think open stack is is doing great in that in that space and that's why it makes so much more sense We avoid vendor locking. We can still create relationships specific vendors to have the stability Um, but we can also ensure that through the open source aspect of it We have a standard that works across all vendors And non commodity I find the the answer interesting. So I do feel we're In a way commoditized on x86. There are still brand names and slas attached to it. Um But I think we've already done the first step of virtualizing a lot of the it and that's coming to the NFV as well So in a way, it's it's commoditized at the hypervisor layer Which should drill down to um to whatever is underneath later I can't add much more but I agree very much with the the standardization aspects of the open source Uh path and I think it'll enable a lot more Commoditization if not capabilities through that So I think and we as Verizon intend to support that and We'll probably try to contribute back into the source environment For whatever special functions. We think we need we'll try to drive that back into the community Um on the commodity hardware again, I there isn't much today Although I can see Very quickly that we are moving in that direction And it it won't be long before a large part of our environment is on commodity environment both hardware and software environments Okay, and so I guess we're gonna change all the standard community to open source community So now with the push of the vendors. Okay, we'll come to the end. So I I'll just ask if anyone have a question There's a two mic here um If not, let's give a round of applause to the panelists And thank you guys for here Thank you