 Okay, everybody. Hello, folks. So my name is Alessandro Pilotti. See you at cloud-based solutions. And i'm peter puliot. I work on open stack at microsoft. Okay, so for those of you who are unaware microsoft has been Involved with open stack since 2010. And we've worked together with cloud-based solutions since Early in 2012, bringing all the functionality of microsoft Products into open stack today. So that being said, we're here today to talk about What new things we've brought to open stack for this release. Yeah, so here's a quick agenda. Okay, so we're talking about windows as a guest. The session is marked as a beginner, so we'll try to Really like to also showing you, let's say, which are the Big benefits of running windows guest in open stack. How many of you guys are running windows guest in open stack at the moment? There we go. Great. Then windows licensing in open stack. How many of you guys understand the windows licensing in open stack? No hands. I don't understand why. Then we'll have someone to talk about today. Okay, then we have good news for you. Open with reach for hyper-v 2.6, so this is one of the Great releases that we have in this time frame. Okay, not technically speaking, strictly speaking In an open stack project, but very correlated. Hyper-v server 2016, as you might be aware, Windows R2 testing just got released. Okay, so we'll see what comes new there, Especially for the freshly released Newton version of OpenStack. Nano server, shielded VMs, One of the great features that we have in Newton, And remote desktop and VDI in open stack. Any VDI users here by any chance? Good. So windows as a guest. Windows can be consumed as a guest on any hypervisor Within open stack. Okay, and there are no differences In the workflow in terms of how it would interop Or function with the open stack services Used during guest initialization. And also, you know, as part of this, Those windows images are typically sysprepped. We've created with cloud-based some technology That helps to, you know, create and master, Go master those images for you. As part of this, one thing you should be aware When running Windows as a guest on any other hypervisor Other than Hyper-V, it's required that you use A certified virtualization driver stack. In the case of using KVM today, You have to use it with an enterprise Linux distribution That supplies those Verdi O drivers. And typically that would be either SUSE, Canonical or Red Hat, And an enterprise-supported Linux distribution underneath. One great advantage, if you have Hyper-V, Of course, all the drivers are already baked Into the Windows instances. While if you use Linux on Hyper-V, Linux was also great on Hyper-V Even in the open stack context. And in that case, they're called LIS drivers. Also part of the upstream Linux kernel. Cloud-based in it. You know, when you have a cloud-like open stack, You need also to have a provisioning agent Inside of your guest operating systems. In Linux, this is famously cloud-init. On Windows, this is called cloud-based in it. So those are two different projects. The reason why there are two different products Is that the original codebase of cloud-init Was very Linux-specific. So we started a completely different project at a time. But the two projects, or the cloud-init for Linux, Let's say, and cloud-based in it, They're merging into a single project. They're called, informally, cloud-init V2. Feature-wise, what does actually cloud-based need to do for you? Well, basically, all the things that you will expect From a guest provisioning agent. So for example, setting the host name, creating users, Setting the user password, static networking, Public keys, automatic volume expansion. So for example, if you have an image which is, I don't know, Six gigabytes, and you have a flavor which is 40 gigabytes, When you boot the instance, you want, of course, To have the volumes automatically expanded. So this is what it's called grow-root, for example, On the Linux cloud-init. WinRM HGPS listeners. WinRM is very roughly the equivalent of SSH in Windows. Licensing. It will automatically activate for you your Windows instances, If you want it. Independently of what licensing model you're using, KMS or not. NTP and MTU. In OpenStack, typically, the DHCP service will offer you A DHCP option that the client can consume to get the MTU size, And the MTP service to be used by your guest list. Windows ignores completely those settings, So what we're doing in OpenStack is that we have a DHCP client Which fetches actually that information, And applies it to the guest list itself. This is very important, especially when you use OpenVswitch, Because you cannot have an MTU larger than what the underlying Tunneling implementation allows. Otherwise, you will have a lot of packet drops. And the most important one I left it for the last one, Which is the user data. It means whatever script you want to run as part of your provisioning. This is typically a PowerShell script, But it can be something way more complex, like a hit template. So all the complex hit templates that we'll see later, Or a Juju Charms or whatever else, Are based on the fact that cloud-based init Can run actually a custom set of scripts, As part of the provisioning process itself. More details. You have the read the docs documentation there. Please go there and get all the information. How do we deploy it? Well, there is a very simple installer, That you simply deploy automatically or manually, As you guys prefer. And when you're done, it will simply get you ready For, for example, c-spracping the image and rebooting. More information about the type of cloud supported, So it's not only OpenStack, It's also Amazon EC2 format, CloudStack, OpenNabula, Ubuntu Mass, we use it very heavily, For example, for bare metal deployment with Mass, And so on. And there are actually quite some more than that. Some notes on the... So we currently supported Windows versions, Our Windows 7, 8, 8.1, 10, both x86 and x64. From Windows Server perspective, We have 2008, 2008 R2, 2012, 2012 R2, x64. And soon to come, Windows... Well, we have support also for Windows Server 2016 today, As well as Nano Server. And it will work on XP in 2003, But it's currently unsupported. Yeah, quite a few people are asking us, Beside, of course, the big question, which is why? Beside this, people say, Hey, we have a lot of legacy with XP 2003. We understand that it's not supported by Microsoft. It's not even supported by us, Meaning that we don't run continuous integration For XP in 2003. So things might break. But there is no reason, actually, For you not to try it if you want it. And if you need commercial support, You can still actually refer to us. Just don't expect us to run Integration services against... Integration testing, sorry, against it. Since we're here, another important thing, Cloud-based innate is fully integration tested. We will talk later about CI in general, But just that you know, Especially with the huge size And installation out there, okay? We want to make sure that whenever We reuse a new version of Cloud-based innate, We have everything tested On every possible release. So just one more second, as a result of this And a result of our testing of this, We've actually been told by some of the Linux vendors That currently the best supported guest on OpenStack Is Windows Server 2012 R2. And actually, if you go look At what the other OpenStack vendors Actually support from an operating system From Cloud, you can do the math yourself And see which one's the best supported operating system Guests on top of those platforms. Hint, hint. Okay. Another popular thing is this Trial Windows Server 2012 R2 Image for OpenStack, okay? So if you just want to see how Windows can run in an OpenStack context, You can just go on our website And download a pre-packaged version Of Windows Server 2012 R2 Readymade for OpenStack. So this is actually in a BHD format if you're running Hyper-V Or in QCOW 2 format if you're running For example KVM, okay? That comes already between Show your drivers and whatever is needed. So CloudBaseinit is there. It's already sysprepped and so on. As a point to make, in general Most people come to us after they've started On the path of using Windows Guests on top of KVM And have problems mastering Those actual images For their deployment. That's usually when we tend to Show them the images that We create, the curation process And the automation that we put in place To put CI and CD around this as well. Just to be very clear The license that you have to accept Before downloading the image Which comes directly from Microsoft Doesn't allow you to use it In production environments. So this is an evaluation only for testing. Images, so in Microsoft Or for Windows at the moment is not Possible to redistribute ready-made Appliances, okay virtual appliance So you have to build your own image. We build images for lots of our customers Of course, but you can do it also yourself. How to build an image? Here there are a bunch of the tools That we use actually internally That you can use yourself as well. We got a lot of questions so At the end of the deck there is also Information about where to have The most important, please refer to those ones. Now from a licensing Perspective, we often get asked On how can we use windows Like how do we adjust our windows Licensing for OpenStack? Or what's the best license to use For OpenStack? In a lot of cases people also want to use The licenses they already have. I guess the best answer for that Is you can use any license in OpenStack However depending upon your use case There are better licenses to choose Then to drive people towards using A windows data center license And if you're going to be hosting Other people's things on it Then it's going to actually have to be The SPLA, sorry Service provider license. So if you have volume licensing Regular licensing keys, KMS All those things can be Used as appropriate licensing Mechanisms in your OpenStack Deployment. However if you're going to be a service provider It's not necessary for you to use SPLA licensing. Any service provider here? Okay, cool. So one of the other big issues that we Have, specifically with Our customers, who tend to Virtualize windows on top of their KVM OpenStack deployment Is avert.io And as we said, the Paravirtualized Device Driver Layer. Microsoft's put in place a program Called SVVP, the server Virtualization Validation Program, which is a mechanism that we Use to certify Workloads on top of Microsoft's Hyper-V. In the case of this Canonical, Red Hat And SUSE All validate their platforms on top Of Hyper-V to ensure That their Vert.io layer functions properly On top of our Hypervisor Stack. So once again, we often get The question, does Microsoft support Hyper-V In OpenStack? Yes, we most definitely Do. Microsoft will support Your use of Hyper-V Regardless of what management platform You use. So If you want to use Hyper-V with OpenStack We will definitely Help you use Hyper-V. If you have questions And you want answers from Microsoft On this stuff, please Email openstackatmicrosoft.com And I will be more than Happy to answer your question Or help direct you to the appropriate person Who will answer your question. Yeah, that's a very important Topic actually, thanks for clarifying, Peter Because a lot of time customers come And say, but it's actually Microsoft Supporting this effort. And the reality is actually This one, I mean we are using Public APIs In our OpenStack development So it's exactly the same as any other Hyper-V deployment from that perspective. All right. Okay, so Infrastructure as a service is great, right? So you get your virtual machines Or your bare metal machines up running Or your containers also. But you might want to get also beyond that. Okay, so beyond going towards Orchestration and everything. One of the Options is heat. So heat is part Of the OpenStack foundation family To say so. So we have Also full support for Windows workloads In heat. In particular Active directory, exchange SharePoint, SQL server Including SQL server always on IIS and any type of Popular to say type of Workload that you might run On top of Windows, okay. Again, cloud business It takes care of running Those scripts, but the heat templates Themselves are actually containing All the PowerShell bits and everything. It's not particularly different Than writing heat templates For Linux with the difference of course That you have to use PowerShell instead of that. Lots of people still think about Windows as like this thing in which You have to click to actually Real Windows in buttons and Visits and all that stuff. That was a long time ago. I would say if anybody thinks That we can't run Windows Identical to running Linux In an OpenStack environment, come See us after and we'll show you. We believe that we can scientifically Prove that Windows just runs Linux in an OpenStack environment. Hyper-V. Any Hyper-V users? Nice. Good. Thank you. Okay, why Hyper-V in OpenStack? Okay, and just a little bit of history here. So, we are The ninth releases, so we start To do it falsely, okay. Today actually, you know When Facebook reminds you about What happened a few years ago and all those Terrible things that come up. From the first summit that we did In San Diego. So, that was October 2012, I believe. I mean, it was amazing. Looking at how our boot looks like today. With all the crowds around and how It was back then when we just started. Nice memories. Well, I go back to Boston so my Memories are even crazier than yours. So, today We have support for Hyper-V 2012, 2012 R2 And 2016, okay. 2016 just got released, but we have I mean, already a pretty long history. Meaning that we supported all the various Intermediate technical Previews and now finally is RTM. So, feel free to go and use it As much as you want, meaning that Since we tested it across all those Intermediate releases, it's already very stable. So, don't think about OpenStack Or Windows 7 2016 as a kind of V1 release, okay. And just to mention We also support Hyper-V on the guest Platform as well. So, when we talk about Hyper-V, we talk about it as the feature That gets enabled on the Windows platform. Yeah, and Correct. So, for Developers, if you have like Windows 8, Windows 8.1 or Windows 10, that's Actually a great idea. We have even a project Which is not meant for production But just for testing, okay. That will deploy OpenStack on whatever Ramps Hyper-V including for example Your laptop with, I don't know 8 Kicks of RAM or whatever, okay. It will deploy all the Linux components in A VM, set up all the connection between The host and the VM and deploy the Host components on your laptop itself, Okay. Very nice if you just want a cloud Inside of your laptop with a minimum effort. Another important thing, so the Newton release got released today, Okay. So, it's live Right now on our website in case you Would like to download it and try it out. Now, CI, continuous integration. How many people are aware that Microsoft Runs one of the largest continuous Infrastructures in all of OpenStack? Nice. Glad you know it. So, we currently have Roughly under a thousand nodes Sitting in a data center in Cambridge, Massachusetts, and we've actually been The number one voting CI since what, Atlanta? Yeah, so Yeah, if you look at stack analytics And you look at the CI votes, you will see That we are basically the second one After the community CI. Yeah, we take Great pride in assuring that our Continuous Integration Infrastructure To support all the effort that's going On here. From our customer's perspective We look at it as a mechanism to allow Us to give our customers a sense of Stability and Higher level Let's say support Given the current situation that we Currently have to deal with. So from That perspective, you know, we try To ensure that all the components We have, well, we do ensure That all the components that we have Are having CI to support them Customers in both our upstream and Downstream releases So, you know, I encourage you to Submit code and watch it Pump through our CI. Yeah Yeah, that's a great feature. You have to consider in particular that open stack Comes culturally speaking from A Linux background and in the early Days when we were developing There were a lot of situations in which Somebody was actually submitting code That, you know, was tested For example, originally only on the Linux Integration CI And it was breaking On Windows, okay. Think about I don't know, Linux signals Or Linux specific code, you know, that got Open stack is written in Python, meaning That can run on Windows, Linux or whatever But of course there are also specific Operating system pieces in there, okay So having a CI is Running on Windows is extremely important Because this will avoid Actually, it avoids currently Regressions when people commit code To work on Windows, especially Not because people do it on purpose Like all those bastards now they will see, right It's mostly like, oh, I didn't know That that specific thing was not supported There, I don't know, the FNCTL library Or whatever else So one more thing to add here Just to talk about what's to come In terms of our CI, we've recently Done a serious upgrade to all The infrastructure and we're about to go live On Microsoft Gen 1 open compute hardware So that's Windows running on open compute And running open stack So from that perspective We have a very good story To tell in regards to Openness and how we're working In these environments, so if you have questions after Please feel free to come by and ask Another interesting challenge That we had was this one Okay, great, we have Hyper-V We have an extremely easy way to set it up And include it in any Open stack cloud, so our goal From day one was always to make Sure that you could take Hyper-V box Put it in an existing Open stack deployment, doesn't matter If what vendor of Open stack, I mean it's creating Your own deployment, your Open stack Or if you deploy it by yourself, doesn't really matter We are used to the fact that In a lot of cases people are adding Hyper-V together with KVM nodes To existing environments, so that's perfectly fine Of course you could deploy a Hyper-V Open stack cloud from scratch But we always have to consider Add it to existing nodes, okay So once you do that, and you have For example, 50 KVM nodes and 50 Hyper-V nodes, you might want To have your tenants being able To spin up machines, for example Linux ones on KVM and Windows ones on Hyper-V And having them talk among each other, no So the first Networking component that we Wrote was the Networking Hyper-V ML2 agent, which is still Of course available and in production It supports VLAN and MVGRE But lots of our users and customers Are used to OpenFlow So VXLan tunnels If not, STT or Geneve now Or whatever else And they use also to use OpenFlow And OVSTB of course to configure their Networking, no? So one of the big Efforts that we did, and I will say Also very successful one, was actually To port OpenVswitch To Hyper-V and to Windows, okay So this works for Hyper-V virtual machines With Windows containers. And now just To make note on this too, OpenVswitch Recently became a Linux foundation project So we feel that the work that we're Doing actually helped to enable that The driver, so we're talking here Of course of a Windows kernel Driver with on top all The OpenVswitch user space tools Which ported on Windows, okay? So the main idea Is that you have to find the Same identical user space Comal line interfaces you will find On Linux That's why for example for OpenStack the same ML2 Linux agent runs In the same way on Linux and on Windows, okay? So we don't have A separate OBS ML2 agent only for Windows, okay? It's the same one This is actually a work that happens upstream To get with VMware and the rest Of the amazing OBS community, okay? So it's not just a separate part in that It's a community effort That it's part Right now also of the Linux foundation So that's an important news They open with which project Would actually under the umbrella Of the Linux foundation itself So OBS was not only Before the de facto standard SDN solution, now it becomes Even more relevant because of that So if any of you are SDN providers And you're currently building an OBS Compliant SDN solution and you want To work to that SDN solution Come and talk to us after this and we can Show you how to do that One additional thing, so the kernel Driver, the Windows kernel driver is Fully certified on Windows 2012 2012 for 2N 2016, actually It's the first Hyper-V extension To get certified on 2016 And of course also on the clients I mean 8, 8, 1 and 10, okay? But again, we recommend the clients For development purposes, not for Server type of workloads, that's Always. Okay, great interoperability Hyper-V KVM, tunneling, VxLand, GUE, STT and so on It's also compatible With Open Daylight, NSX And we're working of course also now With the OVN controller So again, same identical experience Okay, just a quick Note on what's new In Open-V Switch 2.6 Which is actually, you know The 2.6 release just got released Out of stream We have connection tracking, this was A feature that we needed to add Some time, actually, we Scheduled it after the 2.5 2.5 was basically the first Non-experimental, meaning fully Stable release, and 2.6 is Basically the first one in which we Adding also other stuff Connection tracking is needed Mostly for open stack security Groups in this context, okay? So with 2.5, we were basically So the regular Open-V Switch Work for all the Tunneling and so on configuration Segmentation, while we had A separate one for For the security groups Which was actually leveraging The networking hyper-redriver That we use also without OVS, no? Great news with 2.6, we have One single agent to rule them All, meaning that we use Contracting also for Security groups, this also Means that this scales much better Because we just use open flow for everything No ACS managed via WMI Same identical open flow rules That you will find, of course On Linux and so on Additionally, our security features We have name pipes support This is actually very important Because Open-V Switch Delegates to the transport layer The security for the communication Between clients and servers, okay? So with name pipes, we have The same feature that you will expect That, for example, on Linux With POSIX sockets, okay? And we have also a ton Of performance improvements. We are actually extremely happy About this release. Now, so next How many people know what Nano server is? Great, so Windows 2016 Introduces a new concept Of a Windows platform called R which is a sub-600 meg Instance Or image of Severely stripped down Windows platform Ideal for using As a hypervisor or a certain Specific workload use cases One of the things previously, our target Platform for OpenStack was Like Hyper-V server and Windows Server Core. Nano server Gives us a really interesting Footprint To deliver that same functionality And it allows us to Start getting into Sort of more, let's say Typical Deployment models, typical Yeah, I guess layouts For platforms that you would use In a cloud. So extremely small We've done the work to tie it into OpenStack To tie it in with certain DevOps tooling It's, you know, Boots extremely fast and it's Part of the Windows Server 2016 Platform. So as long as You're using that, you can actually Build nano-images And then use those nano-images as your Hyper-V platform in your OpenStack deployment Or as your guest. Next topic, storage spaces direct Anybody using hyper-converged Deployment? So storage spaces direct is a Very important Component, let's say, of The hyper-converged architecture that I'm going to introduce with the next slide It's a shared nothing storage Model. So the main idea is that We're moving away from situations in which You have to have dedicated storage nodes Or dedicated storage appliance We still support them, of course In Hyper-V for OpenStack But we're trying to get to a point in which Basically each node does everything By everything, in this case, I mean Of course, compute, storage and networking In storage spaces, the right comes With Windows Server 2016 And we tested it throughout All the development cycle Of 2016, so including all the technical Preview, and we are extremely happy About the results Supports data mirroring Scale-out file servers for the same Scale-out file server architecture Based on the great SMB3 technologies Available here as well Recommendation, use always SMB director And DMA, slash DMA Enhanced Necks in order to Optimize, of course, the performance And make sure also to calibrate properly The traffic that the storage will require The good thing is that the same Requirements as Recommendation that commonly, say in general For storage space direct, apply also For OpenStack, okay, because again OpenStack is leveraging this underlying layer Here's the typical architecture Of an OpenStack hyper-converged Deployment, so you can see the Four nodes which are fully Symmetrical in terms of components deployed On top of it, you have The Scale-out file server And then on top of it, you have all Your controllers and whatever you might need As you can see behind, okay This might be a pretty Large beast to deploy, because you have Of course a large of components to do it We personally do it typically with Juju Charms and That we develop for this purpose Along with Ubuntu Mass for The bare metal deployment, but you can Also do it with Whatever you prefer, Puppet, Chef, Ansible Or whatever other option, or DSC Okay, we already covered a little bit This part, okay I'm talking a little bit about Fillover clustering, one thing that Came Very often From people coming to our booth Or in general customers coming, you know Peter, yeah, the people, you know One of the typical legacy workloads That people continue to use are Traditional Windows or Hyper-V clustering So we've decided to Enable that as part of Nova, so that way someone can consume Or deploy a Hyper-V cluster In an open stack environment And have it handle The HA component seamlessly Underneath Yeah, the typical thing that happens, you know Everybody's familiar, I think, with the Pat's and Kettle Dilemma, to say so Open stack was developed With Kettle in mind, so with The situation in which you want your guest Applications to handle Fillover by themselves, and have Availability by themselves, right But lots of customers come With legacy applications That they want to move to a modern cloud Environment, coming maybe from all their Type of virtualization solutions Or even from bare metal, you know While at the same time developing Their next generation of applications Okay, we cannot just, you know Switch architectural model and say 10 years to develop a line of business Application and now, well, since I Just moving to a new layer I just rewrite it completely because You know, now we do everything with Microservices and so on, so all those People say, well, I have this large number Of applications, you know, it could be All the mail servers, databases And everything, which don't really fit In the modern open stack model So what we do here is to allow Also Pat's into open stack By using and leveraging At the host level, Hyper-V failover Clustering, okay, which is a feature That enterprise customers are very Familiar with, so it's a very Enterprise oriented customer What does this mean? We will do that to the next slide It means that For example, you can have An unplanned failover You have one of your Hyper-V nodes That just crashes out of the blue Last power or whatever happens So your machines will Automatically be respawned On different nodes, okay So the customer, beside of course The minimum interaction Required for the machine to restart Will simply have their workloads working Okay, if we do it on a regular Open stack deployment, this won't work The node goes down and also all the Machines will go down, so this Means that it will need a manual Intervention to restart You can also have, of course, a Planned failover in which you do Meaning, let's say that you want to Do a planned upgrade of Your underlying host or For example, security updates That will require a reboot And stuff like that, okay So all those things can be Easily handled by the cluster And then the machines can Simply fall back, okay The Hyper-V and our computer Driver updates the status on the Instance, so we directly talk Of course with Nova We have an intention here to have a Second degree, let's say scheduler The Nova scheduler is the only scheduler That really matters, okay The failover will simply move the nodes And update Nova is strictly necessary It will never do it because of Whatever reasons, okay Network events occur like In the live migration use case By the way, the Hyper-V driver Is probably the driver with the Easiest option for live migration Because it comes out of the box And that also leverages as part Of the cluster store And we also test all that in continuous integration To ensure that it's going to work as best as possible When you deploy it in production So try it, it's something that Again comes out relatively often And now for something completely different Shoulder VMs Shoulder VMs are something That You know When we talk about hypervisors In 2016, we usually think About this Layer in our infrastructure We got completely commoditized So at this point, we don't necessarily care If it's a VMware, if it's Hyper-V If it's KVM and everything Quite different from I don't know Ten years ago in which the features That were coming with hypervisors Were actually the reason why people Were choosing one on top of the other No doubt there are a million of the reasons Why people would prefer One or the other or maybe For example OpenStack compared to something else Shoulder VM It's one of those features Which only Hyper-V has So all this Preamble just to tell you That it's not necessarily true That all the hypervisors are the same Each hypervisor has also different features So Shoulder VM is one of them What's the reason behind Shoulder VMs What happens if somebody Takes control of your host Normally It means that they own All of your VMs Shoulder VMs address Exactly this situation Your underlying host Even if it gets tainted So if somebody gets control They won't be able to Access in any way the virtual machine Running on that So this requires a given set Of components where you have A host guardian service For example the attestation And key protection and everything You have the guarded hosts Which are again guarded by the guardian As you can imagine And then you have the shielded VMs The shielded VMs are just regular virtual machines Which are running with all these additional features For example the disks are completely encrypted With locker Tempering is avoided So you don't have any problem With the host Even if the host gets compromised Or financing or situation In which you cannot allow Something like this to happen And again it's just part of Windows server Line migration is fully encrypted There is no console access So nobody can access the machine directly Unless they do it remotely via network Meaning passing the security The security groups Rules that you might have Instances are booted Actually by a shielded template Which is basically just a glance image There you can find the link For more information Which I strongly recommend Here it's how it works You just upload the template In glance and you just Set a bunch of properties So for example you have to say That the glance image is shielded That you need to have a VTPM So we have virtual TPMs today Secureboot has to be required And a PDK that We don't have necessarily the time But it's a say one additional Info that the user has to pass In order to boot the image We use by the way Barbican Another great Opus Tech service To host all the secrets In this particular deployment There is a link to a video That I recommend you to watch And I will skip here Because this is a 40 minute session So we better add some more content here So one of the other typical Use cases that we see For Windows and Hyper-V And OpenStack is for VDI And we've Worked with one of our customers recently To sort of Let's say develop a solution To use OpenStack As the ideal Sort of back end for your VDI deployment Alright So obviously most VDI done today Is on Windows guests And we do believe that OpenStack Is layer to offer VDI solutions on top of The VDI solution that we're Currently working on Enables both Microsoft RDS And Citrix Zen desktop And can be also integrated With other app layering technology Like Unidesk So one great thing about VDI is the fact that When we talk about virtual desktop infrastructure We talk typically about OpenStack as one of those guests So OpenStack is the perfect Infrastructure service back end For that What typically misses Is the layer that connects The underlying infrastructure service With the provisioning And the brokering And the way in which the users Can actually consume those desktops You can just take a Windows instance Deploy it in And simply have access to That image But that's not exactly how VDI works VDI is much more than that You need to have also policies Of course brokering You need to have an HTML5 type Of access to it So that's actually what we did And also app layering For example Unidesk So here it's a diagram Showing how this thing works So you can see that you have Connecting to a gateway So the gateway will actually provide A first layer of filtering and policy And this is actually the part Of the Microsoft RDS infrastructure So the same identical RDS that you know From Windows Server Here is applied in the OpenStack context From there we go To an IS front end That will basically Serve the applications and desktops Behind there is a broker And here is the most important piece In the broker we have a plugin That instead of Using directly Hyper-V boxes Or Windows servers Will actually deploy the individual Machines in OpenStack So all the logic for the broker Is inside of this plugin that we wrote It works great and a lot of times People are Let's say that RDS is great But it's limited in how much You can scale Well with OpenStack It's actually scale is great As much as you want So this is actually a great solution And behind as you can see You have all the various OpenStack zones So what happens after that plugin Is just straight OpenStack So you get basically the best of the two worlds This means that you can use Whatever hypervisor you want It can be KVM, it can be ASXI And it can be of course Hyper-V Hyper-V has an extremely important Advantage That is designed for running Windows on Hyper-V itself And RemoteFX is one of the great advantages that you have So RemoteFX got also Improved with 2016 And it's fully supported in OpenStack, okay In the OpenStack driver I think we're getting more or less at the end Just a note Huge performance improvement Since Mitaka and also Newton You will see pretty soon A blog post in which we'll compare Performance within Hyper-V and KVM And other hypervisors And prepare to be extremely surprised About what we will see It will be based on rally And we will open source all the scripts So anybody can retry them Just a quick note on containers So Windows Server 2016 now supports Windows containers So you can use Anyone familiar with Docker Experience when interacting with Windows And as you can imagine We decided to take all that Great Windows container technology And integrate it into OpenStack To ensure that all the existing Container technology inside Of OpenStack can now consume Windows containers In that use case So they're currently working On adding support for Kubernetes Works with both Currently on the 2016 platform Windows containers There's a Hyper-V container And a Windows container The Hyper-V containers Typically don't they add a different networking Security layer on top of the Container platform So depending upon the use case You would determine which type of container You would consume And I believe we can also This allows us Can we run Hyper-V in a container yet? I can't remember You can use containers With Windows You can use Windows containers On OpenStack So if you have any questions about Windows containers drop us at our boot We can show you, talk to you and so on Last thing about We have full due to support Charms for all the possible Windows type of workloads Here are some examples that you can see From Active Directory, SQL Server SQL Server always known so clustering SharePoint, take change, you name it Okay So let me end the Session inviting you to our Boot which is A5 We have HoloLens And we have a demo Showing you how to deploy Windows, sorry Windows How to deploy OpenStack VMs In this virtual world Okay, last thing Context, so OpenStack at Microsoft.com It's regarding OpenStack And if you want to hear it from a Microsoft person You can email that address And someone will get back to you Okay, and then we have also Ask at cloudbase.it And of course they ask platform from OpenStack I think we are at the end of the session, right? I think so, any quick questions While we're running out of time? No one, go for it So specifically using what Can you give us more context? We do use Windows as a host Please, sorry, because it's taken So the question is What about network Speed, because You see when you have Any overhead On the infrastructure It results in Network overhead Even you may have problems in your login System if this is a RC Slog Have you experienced something with that? So, you know, we obviously Use it, you know, depending on what Type of encapsulation you're using You know, hardware acceleration is Pretty much required For certain, you know, with Deployments You know, typically I think I'm not sure, I know you've done more of the Performance testing Well, of course, encapsulation Adds a layer of complexity, right? And that's not Hyper-V specific, it's any type Of situation in which you do that I usually recommend to use VLANs When it's possible, but it's limited Of course to 4,000 and 95 tenants, right? But For VXLAN, GRE and everything What we do is to use Hardware acceleration and hardware Floating when it's available And limiting, of course, to the minimum The amount of work that needs to be done That's of course not window specific It applies to KVM with Linux It applies to Hyper-V or whatever else So it's a Generic design Question that you have to consider But I guess if you have something specific Please feel free to come talk with us No, thank you I just know about some Comparative disadvantages In Windows Okay, come to speak with us Or to the booth, okay? I can tell you That there are no competitors Yeah, we like hearing that Because we want to disprove you Okay, thank you guys Anyway, thank you everybody