 He runs two amazing user groups at least two Yeah So I'm not gonna steal your thunder you're welcome to talk about that as well And I know you have a dynamite presentation coming up. So I'm just gonna pass this on to you. Thank you so much Hi everyone, I'm Ken Hoy so I've Been involved with open stack for a couple of years. I'm currently the co-organizer for Open-stack user groups in New York City, Connecticut, and Philadelphia Co-organizer so so we've kind of created a model where Since I know a lot of folks in the community my job is to arrange speakers and then in each location we actually have a local organizer who's in charge of logistics and facilities and Because as you can imagine New York is a big draw for a lot of vendors in terms of wanting to participate We've kind of adopted a model that basically says if you want to come to speak at the New York City meet-up You have to agree to speak in Philadelphia and Connecticut that same week And by doing it that way so it's a model that would sort of testing out which would be the idea of You know one way could spread a lot of user groups where in small areas where maybe vendors aren't as incentivized to speak would be to create a create a anchor It's not an anchor user group in a large City where vendors want to speak and then built a lot of smaller user groups around it and basically say you want to speak of that Large user group you have to speak of the other smaller ones as well So that's a model. That's works for the most part for us. So So I actually try to be fair to the Vendors so if you guys know any about geography Connecticut's New York City's in the center Connecticut's North Philadelphia South So I try to make them go out of north to south or south of so New York City's ends up in the middle So it's worked out. So that's one of the things that one of the ideas I'm trying to float by made the foundation Speaking of which that there's I'm also part of the Open and open stack ambassadors group. How many you're aware of that group? Only a few of you. So it's basically we were Selected by the foundation as folks who've been involved in the community and part of my charter being part of that group is we are To spread the word about open stack and to help people who want to start user groups so if Obviously, I'm best capable of helping you if you happen to be near New York the East Coast But doesn't matter if you're in another country if you just need some advice I'm certainly happy to help or I can point you to the open stack ambassador who's covering your region So that's why I say about these groups What I want to do today is go through a presentation So there's a talk I've actually done in the past in all the summits About open stack for VMware and the story behind that really is I actually have a lot of experience with VMware And when I got into open stack was around this Shortly after that VMware started making a lot of noise about being integrated into open stack And I found there wasn't a lot of documentation about that So I started drilling down on the existing documentation and basically bugging the Open stack team at VMware as much as possible to find out exactly how things actually work So I've kind of given this presentation And a lot of the focus is this also is that I used to work at rack space on the open stack team And I would say half my Conversations with customers were trying to talk them out of using open stack And why I mean by that and you'll see is there are certain things like Oracle rack That may not be a good fit for an open stack cloud and sometimes customers on users aren't clear on that So someone walk through that in this presentation Okay, so why VMware and why and or use open stack with it? That's a great that's a great place to start because I think there is a tendency for some of us Who like to chase after shiny new things and you go open stack to shiny new thing and oh, it's free So therefore we will use it to replace everything. We're doing the VMware I'm going to tell you why that's not a great idea Because one of the the key thing that you need to think about in any kind of Architecture design infrastructure wise it's that the dictum is workload to take dick takes architecture So you don't select your cloud platform and then say let me let me just throw as all my workloads on there You start with the workload you start figuring out what it is. I will actually want to run What is it that my business needs to accomplish their goals and then say I will find the architecture the platform that fits Fits underneath that workload. Otherwise you will have Failures and and your users will not be very happy So we talked on how many of you familiar this term third platform second platform. Have you heard this before? This is kind of new so just think of a second platform is traditional IT workloads like Oracle ERP Right exchange perhaps their platform are things like Twitter Facebook kind of web scale Very social, you know a good example is I have a customer who's a who's a big bank Right and they've ran almost all their stuff on you know mainframes even but certainly also on like big iron You know Oracle applique running on HP servers go on a you know three par or or an EMC, right? but That when they did that they were taught they were really only focused on getting data out to these ATMs, right? You know Which is could be large, but it's not super large and now all a sudden they're going You know what? I need to provide the same level of services and data To users who may be accessing that data from five different devices Right from an ATM to a PC to their phone and all of a sudden that the Exponential amount of growth in that in in that scale means that I can't use that same legacy application So this bank is looking at things like object storage for example To store your check images, right as opposed to storing on a NFS file system, which can't scale to the levels that they expect they have to scale So that's that's so that's that's an important differentiation between what is a second platform workload? And what is that third platform workload because knowing which one you're targeting helps you decide which architecture which plow platform you need to run on? So if you guys are traditional kind of you know data center folks You'll recognize all the second platform stuff, right? These are these are things are very where they the application and the OS and the hardware is very tightly coupled, right? If if you're running Oracle on a server And it's a sands, you know Sand storage if that storage dies your oracles going down is going down hard, right? And you and now you're in a failure recovery scenario The other things that these tends to be very much a traditional Database servers again like an Oracle or SQL server that's certain that assumes that the infrastructure on it Underneath it. It's always up and running like kind of 90, you know five nines availability, right? And it doesn't deal well the infrastructure kind of goes away all of a sudden And they tend to be steady state invade so I know that if some of you are working in places where they are There's such things of peaks, right? Maybe at the end of the month or the end of quarter But generally speaking with traditional apps, you kind of know what the boundaries are, right? You every month, you know, you're gonna spike up by x percentage more or less So you can kind of built, you know, you ever heard that kind of built for the built for the peak, right? Because it's because you're not gonna have a very elastic infrastructure You want to make sure you can account for that, but you know what that peak is at the end of the month or quarter And then operate a focus. So what how many you guys how many of you here have done work on things like? Windows or all of VMware for example Few of you so I always ask this question when I talk to VMware admins, which is how many of you who are doing VMware admin work? Give credentials to your developers to the web to the web for your client and say you go ahead and provision your own VMs and storage So I've talked to several hundred. I found one guy who does that, right? Because that traditional workload because it's of its complexity, right? It assumes that you don't want it the end user to have access directly to that resource, right? So the cloud so the admin becomes kind of the gateway to provision resources and As a result you need a very resilient type of infrastructure, right? Could be like a V block but it could also be HP's cloud system, whatever it may be IBM's Where the go the go here is the storage never fails, right? Well, it's because it's highly redundant or you have depending on something like VMware, right? Where you assume when the when the VM underneath when the sorry the hypervisor Server underneath the workload fails all the VMs just restart and the application shouldn't even know is running on a brand new physical server, right? That's in direct contrast to these third platform There's a web-based type of workloads that needs to scale right in these in this in this world Applications don't necessarily have a one-to-one correspondence sometimes an application lives across multiple serves Right, and you could have a piece of that fail and the application will continue running and just recover itself somewhere else And these tend to be very unpredictable, right? It's kind of it's the type of workloads where You know you put out a new product on the web and you you're sizing it for a thousand users And actually it's possible to spike up to two hundred thousand Right in a few minutes and then back down again, and so how do you design for that? Right, how do you if you have to if you have to like provision of V every time you provision of VM? You got to ask an administrator do it for you How do you scale for that isn't possible, right? Everything you have to automate everything which is again one key tenants of of a third platform cloud, which is what I would Put open stack as primary designed for it's very When I say developer focus it means that everything everything is exposed via API's But develop the idea is developer can provision his own Machines right his own storage and networking without the operator having to come into the middle And that's how you can scale very rapidly and the design assumption at that kind of scale level is that? the infrastructure will fail Right, so the app so This is in this area is very important because again when I talk to a lot of banks traditional enterprises They they go can I put Oracle on here on open stack? And I expect that when if a compute no fails although Apple Oracle just recovered by itself The VM that's running it on in those machine And I actually have to tell them no that's not the case Open stack is not designed to do automatic fail over when a VM when the hypervisor no crashes Which I never brings up this question, right? You guys have been at it for years, and you guys can't figure how to do VMH in automatic HA What's going on here, right? Why is that? Why is there's this deficiency in the product and the answer? I give is that you need to know what the design principles all right opens whereas VMware was very much designed for resiliency right on a on a limited I would say fairly limited scale Open stack because of its desire to be a private cloud Alternate like the private cloud like at AWS is designed for rapid scale Right the idea is to grow really fast right and when you do that certain things inevitable right so if you have a VMware environment with a hundred, you know a thousand VMs That's a fairly sizable one, but I maybe that sits on you know, 10th of 20 30 servers With a tied to a sand storage It's a small enough environment that you can design this thing to the infrastructure not fail What happens when you're talking about web scale and now you're talking a couple hundred thousand VMs Right and and it's spread out across thousands of servers Right if you look at the probability is if something will always fail somewhere Right, it's impossible in that large-scale and environment to be able to ensure That infest that the server doesn't fail or the storage or the networking or even the underlying VM Hypervisor will fail something will always fail somewhere So the goal here is you need to just assume that it's a case and built for it So you design for it to fail you assume the wheels are going to come off Any time in the infrastructure and you built you design your applications When you're on OpenStack to be able to account for that. So I'm actually gonna skip this I'm assuming all you know what OpenStack is right. So OpenStack orchestrates a lot virtual resources And presents them up right big again very importantly as self-service on-demand APIs for you and users Again the goal very clearly here is not around infrastructure resiliency the goal of OpenStack to figure out how you can help developers deliver I oh sorry Operators develop a self-service IT to their developers at rapid scale and things like shared Infrastructure sands like you have to run on VMware One attendance is the more shared infrastructure. There is the more limited just scalability Right so by having a non Not nothing shared architecture in OpenStack All right not using a lot of sand disk You're actually able to scale more rapidly than you could in a in a shared sand environment But they never the reality is as we're as OpenStack is maturing and getting into the enterprise, right? More and more customers going why I want to I want to run both Right how many how many enterprises have nothing but platform 3 well That's all they have coming up new no one right everyone's got platform kind of that older platform 2 stuff And and maybe they're growing into that platform 3 and they want to be able to run both in their environment right there is I think there is a desire for customers to say I don't want to have to run two different cloud platforms That are managed completely separately So VMware kind of got into the game kind of the dress some of this and Today, there's actually a couple of ways you can do this right one way which VMware has done a pretty good job of is they basically took vSphere which is their hypervisor technology and folded it in underneath of OpenStack Right because one of the things I shared talked about is OpenStack is actually not a hypervisor, right? It is a orchestration of virtual resources including multiple including hypervisors, so you can actually Run OpenStack with KVM with Zen with with Hyper-V or with vSphere in this case So which so the way and the interesting thing about how VMware does it is they basically take a vSphere cluster right and expose that to the cloud the OpenStack services Nova services as a single compute node So the benefit that is from the note from your API you can provision Machines right to a vSphere hypervisor just like you do now with to KVM This is exact same API commands But from VMware's perspective once once the command goes out to supervision of VM in under VMware It actually uses vCenter So now you get all the stuff we talked about that you can't do with OpenStack generally natively like having VMs fail over automatically or happens underneath vSphere underneath of OpenStack So one of the models that we try to do is basically creating two Availability zones of your favorite the concept of availability zones Basically your court you're gonna section off a subset of your infrastructure and say that Avail that's called my my platform to my VM where Availability zone and your develop can choose to provision VMs and Applications onto that zone and then you can have another zone one in KVM We've all shared storage and you can run all your web applications and you're no SQL databases All right, so that's a that's a good approach to take Second approach how many have heard heard about this VMware integrated OpenStack Some of you have so very quickly. It's just this is VMware's own distribution of OpenStack Right, but it's the concept here is that of a what they call think of it as an under cloud and over cloud So the under cloud it actually is what the VMware admins actually Where you put all your open stack services that actually runs on vSphere and then you have an over cloud Which is your open stack services and like Nova and Neutron and the idea is if I put if I put this on top of that over cloud under cloud now I have a I have all the resiliency that y'all have a vSphere and I can I can manage it like it's a VMware application So some of you guys mentioned your VMware admins. So think about what if you know your boss came to you and said You know, we need to spin up OpenStack for my developers and you're scratching your head because you don't you haven't played with OpenStack Right, and but now you've got a week to get this rolled out You know and some of you have done OpenStack know how complicated that can be What if Doing that means just you basically right click on your VM your web vSphere web client And there's a V app an app icon that says open stack you right click answer a few questions it just basically blows out an open stack environment on top of your VMware infrastructure and now From the developer's fixer is that he's using all the AP OpenStack API's like it's a native OpenStack But for you as an admin you just treat it like it's VMware. Yes Yeah, so right now the first cut is going to be VMware only but they plan to support KVM as well so this is There's you know it's posing counts everything but the big pro here is if you're a VMware guy or You're working VMware guys who need to get open stack up and running This is probably the fastest way because it's the most familiar tools that that they they have So I think my time's up. Let's just it's time for questions or anyone have questions Yeah Yep, so what they use you familiar something called NSX Basically NSX has a has an option called multi hypervisor Basically you run that and you can you can run it through open vSwitch on a KVM side side and then the NSX switch Also, it's NXX on cross both environments Yeah, that's what you could do. Yes Well, so so when I say so basically I would say all the vSphere clusters are its own zone And you and you want to do like Oracle you put it into that zone and it'll make Then Nova will make sure that it only gets provision in a vSphere cluster that provides the HA Obviously there's stuff underneath that like you have to have a share sand, right? Otherwise, it won't work Yeah, yeah Next slide or previous line. Yep. Yep. So this is this is for the So VMware as I mentioned before is a very operator focused So I don't know anyone that loves the VMware API except for the guy who wrote it. I happen to know who is Developers though love of the open stack API, right because they can do everything they need So the idea behind is we're basically cloaking VMware environment Behind these open stack APIs from the developers perspective. They don't know that is VMware They just think it's virtual infrastructure. They provision Right, but you're getting all the good, but the the key thing is the cloud admin Who knows VMware can manage it like is VMware and it can use in fact I'll use all the tools that comes of VM like vCloud to actually monitor the environment Okay Okay, yeah Yeah Yes Yeah Or somebody Both opposing cars right so so Not a civil bullet. I think it's got it's good. There's a great use case for using it But there are certain use cases where Via, you know running native open stack may be a better option depends, right? So So you're bored by I wrote actually two blog posts earlier so way before VMware announced I wrote a blog post. How do you actually work VMware with open stack? So I had to use a couple of the reference architectures all those things. So there you can actually see Yeah, this is the one There are plenty of seats So in this slide if you see I think I wrote it in March 20th So VMware hasn't announced a video thing. We have already did it open say that thing So how VMware actually integrates open stack with other partners like Mirante is canonical red hat and Yes So this is the architecture that you see So if you look at this, so this is this one is with So here on the on the on the violet color you see all the violet boxes are basically open to products Right. So here you see one to Mars, Juju and landscape and charm and here all the blue stuffs are VMware's product and services and all the orange Ones are the open stacks other side So if you look at here in the cloud operation, it's at the vCenter vcox and the login side manager, etc And then there you know vSAN you can connect to vSAN or third-party Storage here, right? So this connects to sender and then you have NSX product and VMware networking So which will connect to your neutron which will manage the neutron and then you have vSX So I'm just thought that this was kind of relevant to the conversation we just had with Kenneth so