 Cool. How you doing guys? David Patterson here with Dell. I have Corp. Presenters Rick Pioso who's also a Dell employee and Akalash from Intel We're going to talk a little bit about the new business unit that I work for at Dell then we're going to get into the Intel product that we're integrating as one of our solutions and then Rick is going to talk about How we are leveraging ironic to do some zero-touch provisioning of edge nodes and then I'm going to summarize and Take it from there. So as I said we're part of a new business unit at Dell where we're Focusing on Where telco and cloud and edge converge, right? The the main offering that we've come out with as a group as we've grown over the year is Dell services edge We'll go through the The smart edge overview and then we're going to talk about Rick's going to talk about DHCP Pixie challenges at the edge and how ironic can help and I'll give you an overview of the solution and then Q&A and references So as I said telco systems business Edge solutions of which I'm the lead is a we're heavily focused on mobility edge based solutions So Dell services edge the first offering we came out with is based on Intel smart edge Red Hat OpenShift is running the smart edge controller and we're using Airspan 5g CBRS radios and litmus edge is the IOT platform We're also moving forward with a variety of other Private mobility solutions that are in early stages of define I have two tracks that we're just finishing up define phase on that will be we'll start working on soon all about adding IOT Machine machine learning digital twinning inferencing and new use cases are being added all the time So one of the big challenges with the group is that we have people that come from the telco world that have you know 20 years of telco experience and know all of these protocols and are familiar with all these Appliances, but they don't necessarily know virtualization containerization You know massive scaling zero tough for the shitting DevOps infrastructure is cold all of this stuff that Cloud developers like myself now and on the other side we have people like me who come from a cloud background and telco a lot of this stuff is alphabet soup All of these standards like 3g pp, which is a combination of a bunch of different organizations Including Etsy Then we have 5g stack which has a bunch of disaggregation flavors We're using a seven two dot split right now for for our solution, but there's split six and Others coming down the road There's also a lot of competing standards they go ran alliance small smell small cell forum Who's responsible for split six? Intel Flex ran and getting my head around Spectrum and some of the terminology are on there like g node b e node b all of this stuff is challenging so Both the team from both of those angles this has been the challenges like we're we're both learning And it's it's you know, it's been a challenge, but it's also been very educational So the convergence of telecommunications in computing It's as we can see the cloud and edge and IOT all this stuff is gonna be massive, right? Edge and 5g will be key enablers Wi-Fi is not gonna go away, but 5g is definitely going to be a Big part of the new next generation of computing Security first and foremost You Use cases that are coming are our legion right there is just Thousands and thousands of things that you can think of doing once you have this kind of connectivity and these kind of compute sources So, you know, you're talking about augmented reality seen intelligence ML ie Autonomous everything And our group's first major offering is an effort by Dell Intel the smart edge product red hat at OpenShift and we're using air span for radio equipment specifically their 5g CBRS and the litmus IOT platform And the current release is still services edge 1.2 He did after Akalash. He's gonna go over smart edge, right? Thanks, David. So I'm Akhilesh. I work for Intel and particularly on the on the product Intel's smart edge So here to put in perspective Intel smart edge is a is is a software solution that runs on any Intel platform and and we are partnering with Dell to enable edge computing solution to also Any customers that that they wanted to apply the infrastructure and in this particular case This is an edge solution for specifically for for telecom applications So here the challenge that we have with with the edge solution is is to deal with multiple software components So it may be with orchestrators Well, maybe the workloads how to how to draw the you know analytics and act on it So in this in this particular solution that what our colleague will be will be showcasing is to deploy the solution and deploy and manage so in our case Well, let's say if we take this product and want to deploy this on any of our existing Enterprise, let's say private customer. So somebody somebody has to manually go boot up an ISO So the solution itself is an ISO the operating system based off of Linux And and also it it works on the principles of Kubernetes to interact with an orchestrator So somebody has to go and put down all these components and configure them Mainly for this telecom solutions like we ran if it if it is NFV you have to make sure that all the networking components are set properly And also since this is a this we have smart edge open and also smart edge commercial solutions So depending on what kind of flavors customer choose Then we have to make sure that when the ISO boots up the node instance You have to register with a controller and you have to manage from a remote controller solve this process the manual process involves in generating tokens and then exchanging with a controller and And make sure the handshake happens, you know, licensing is all met So the in this process. There's a lot of manual intervention So what Dell is trying to do is integrate into the framework with a zero test provisioning using Ironic so that seems like what we where do we where we want to be and You know, this is some of these slides are Just just giving us a software overview of what the smart edge is so under this is the, you know We use the community best solutions in closely integrated with with Kubernetes So the underlying service that we want to deliver is Is to have is to have what in any cloud native deployments that you have in general cloud We want to bring that to the edge and it's also we want to make everything less compact and minimal like any any DevOps engineer or any any IT engineer wants to deal with these components They they want it to be at the very minimal set of any technical interfaces I guess come in with a minimal set of knowledge and And to be able to you derive these features right off of the software and the hardware platform that exists So one of our key If you ask like who is being The primary recipient of this product is is whoever wants this 5g connectivity Because it's the product is meant for edge Customers and whoever wants to use computing at the close proximity of the user has to use Use this in a solution and being edge close customer to the user Which means the user is always having less knowledge about all the management characteristics of the product, right? So I'm going to skip skim through this very quickly. So we just want to emphasize here is that At present the product product does not have Zero test provisioning feature as per se and it has a lot of manual intervention to bring it up to manage the node And one of the gap and also what we are trying to fill from from this effort is to manage the workload which Primarily the ran, you know when when we bring up the ran service We have to bring up all the radio radio software components along with that 5g core So that the end-to-end connectivity can work So the zero-touch provisioning will enable us to bring up the node bring up the workload all of it in one single one single input and then one zero touch so we'll skim through and The other benefit of using smart edge I mean used to have the you know the underlying hardware features Transparency gets exposed to all the software layers. So the orchestration doesn't have to figure out Okay, which which feature exists which doesn't you know which person doesn't it doesn't exist? So we want we want to be transparent so application can when it lands it can know what exists and what not exists So we've also want to you know translate a little bit of burden from the orchestration to the workload itself to they can dynamically switch so just just to make sure The product works best for the edge solution and the edge edge applications and also to be in the ecosystem of cloud And have and derive all the benefits from it. So I'm going to hand over to Rick who is going to have that more technical details of zero-touch provisioning. Yeah, thank you. Of course. Thanks so Like to discuss how ironic can help with zero-touch provisioning and the deployment of the workload So I think what's some most Significant in terms of its ability to assist is that I the ironic community has adopted The modern redfish restful API web technology for managing bare metal the ironic Codebase Offers a number of feature functions and they can be used to zero-touch provision a system Those include inspection, which is typically out of band with the assistance of a baseboard management controller BMC as well as in band introspection by booting typically a RAM disk Also firmware update either day one or day two Can be performed out of band with redfish Burn in so a lot of a lot of Houses a lot of a lot of organizations when they get a new delivery of hardware Before they put it into production. They want to burn it in to ensure that it's stable and reliable and reduce Maintenance headaches later on that can be done through ironic cleaning The next thing that ironic can can assist with is configuring the bios settings Which is not to be confused with the the bios firmware which you might update But actually the various settings that a bios will offer to basically tune your system to your particular workload things like Virtualization is a very common one of course You can also create a raid configuration out of band Or in band for that matter And in the out of band version uses redfish to do that You can assert a boot mode so you could either use legacy bios or you fee and Also, you fee secure boot is a relatively new feature that was added to the ironic code base Of course ironic after deploying the operating system needs to be able to Inform the system the booth from it So throughout the various workflows that ironic supports It's able to set the boot device to use on the next boot then more recently we've added telemetry to the ironic feature set and Of course, there's more So a major feature that is is very useful for this for this edge environment is virtual media boot It's a feature that's offered by redfish It's more secure and reliable than the Traditional method that ironics use which has been pixie and I pixie and it's especially well suited for the unreliable communications environment Or environments that edge deployments often often find themselves in The way it works is that the baseband or baseboard management controller is informed to Insert the virtual media and when it's an informed to do that. It's given a URL To go ahead and visit and get the actual content of the media So it does that and then presents it to the system with the assistance of bios as if it's an actual CD-ROM or a USB stick Fund drive or even a floppy. So to the system. It looks like real physical Like a real physical device and then the system can be asked to boot from it Virtual media boot can be used to boot either RAM disks Which is basically the replacement of the traditional? Pixie or I pixie boot of the ironic Python agent or more recently can actually boot a RAM disk that represents an actual instance Operating system environment, which is how we've actually used it with our our work with the mech and What it's important to note is that this virtual media boot feature is foundational It's required by the other features that I'm going to go ahead and discuss now So the two features that we've used that are built on top of virtual media boot our layer 3 DHCP less booting of a RAM disk and RAM disk deployment So layer 3 DHCP booting of a RAM disk Eliminates the need for both TFTP and DHCP You basically configure a static IP network configuration on the bare metal node object that ironic Manages and that is applied by a cloud in it or green and As I said, it's built on top of virtual media The next one is RAM disk deployment. So RAM disk deployment In that case the ironic deployment It's an ironic deployment approach that does not actually write the instance operating system Onto non-bottle storage like a hard disk or raid volume, but rather just leaves it in RAM and So the the intended work Workloads for this feature are two those that have a need for just ephemeral workloads such as high performance computing and Scientific computing or and this is the way we've used it Is the deployment of an instance OS which in turn? Will write the actual workload to the disk so you boot a RAM disk which then by a virtual media Which then in turn looks at the system figures it out and installs the operating system and any other content that it wants on that system and then That instruct in instructs ironic to boot from that Subsequently into the future So I've gone ahead and broken out the various commands through the staging of what we did There there were four basically force four stages The first was to create a development environment and the development environment that we decided to use was by frost It's a really simple to use tool It's ansible based And includes a bunch of playbooks for installing and executing Well for installing ironic in a standalone mode and then using that installation to execute common work Workflows from an ironic perspective. It was very easy to use. I highly recommend it if you're interested in You know experimenting with ironic and using it in your in your environments the next thing Was to actually create an ironic bare metal node to manage so the first command which is multi-line there is to create the node and The emphasis is that we used one of the two drivers that highly support redfish We use the iDRAC driver, which is used to manage a Dell EMC power edge server there's also a Community-based redfish driver, which is called redfish. They both they both support these features that that I've been describing and The other thing was to support the The RAM disk boot We said force persistent boot device never So it it only boots once from that initial RAM disk that Then looks at the system and deploys the final workload on the disk And then the boot interface was set to iDRAC redfish virtual media that's important of course and then the deploy interface is RAM disk and The final step there is to put it in a manageable state so that you can do the rest of the things that you want to do with the system So the third bullet is about configuring layer 3 and DHCP less for inspection So we used it to inspect the system and find out about its capabilities the first the first step is to set the static network configuration information on the node object and then the second one is to perform the inspection of the system and then finally there's deployment and The first step there is to say please don't automatically clean the system Cleaning can be used if you want, but we we didn't use it in our work It can be used to do things like configure bio settings rate settings burn in those kinds of things and then we provide the node for deployment and We set the source for the for the RAM disk image to be HTTP And we give the location as to where it could be found on the web and then we say deploy it and it goes ahead and deploys So like to hand it back to David. Sure. So the work that Rick did was very useful before We did this integration with with by frost. It was a you know very manual process and in fact typical Smart edge rollouts if they're doing a large deployment second touch So they'll take it ship the boxes to maybe a third party and they they will burn the bits onto the box This will allow us to do it in a more zero-touch provisioning manner and in the latest version of smart edge There's two pieces of smart edge is the edge node itself and then there's the controller Which is running on OpenShift so edge node 5.6. They added a new API call, which is factory Which allows you to pass in an activation token these activation tokens Come directly from Intel and they're you know very secure And then the next thing is they have set it up so that there's a default FQ DN for the controller And you can use that to In your own DNS server to automatically map To a name record that is your your actual FQQ FQ DN for the controller So this takes us that much closer to zero touch for the current version of smart edge These things were just we just got the bits last week. So it's it's brand new and it's it's very useful State diagram. I don't know how clear it is, but it basically narrowed it from there To here. So there's a lot less interaction We do need to still make one call to the discovery service, which is on the web But it will just return the the same constant that I mentioned there in Number one and number two there that one And Yeah, that's what that's what's enabled us to get that much closer to zero touch the one piece that we cannot do yet is The Final enrollment in entitlements, that's the word I was looking for I apologize the entitlements is a manual process Because once you activate the node and register it with the controller entitlements are handled by direct interaction with Intel so The emphasis is all about security with smart edge, so it might sign sound kind of cumbersome, but At the end of the day, it's it's a very secure platform Also includes TPM support and it you know, it's very very secure Sometimes it's actually difficult to work with it's so secure because you can't even really get into the edge node without getting a token from Intel representative so that might sound bad, but on the other side it is it's very secure In our example, we're doing just a single node Deployment of the smart edge control controller just to a VM But in our release, we're deploying the controller to Red Hat OpenShift 4.8 So Intel it's been I don't know about nine months or so that worked with us on Disaggregating all of the services that were deployed via VM into helm charts and containers, so we're the first ones to get a You know Kubernetes based smart edge controller. It was one of the big the big things that we got with this release The edge nodes we use was what I had on hand, which was the power edge XC2420s They are the previous generation of edge nodes and I believe will be end of life pretty soon the actual RA though is Supporting the the next gen XR11 XR12 Which is even shallower than the XC2420 and supports more modern CPUs Also, they can be pre-configured to have front-facing or rear-facing ports so they can support a variety of different environments So the actions we took was we leverage ironic's BERT boot from virtual media for Doing the actual laying down of the operating system And activated the node via the new REST call I mentioned And then we need to imply the entitlements manually And then a enrollment is complete Whereas before in the standard process it was We would install the ISO As well you still attach it to the BMC as virtual media, but it was a manual process Then we would have to get in touch with Intel to get a activation token Then we would go through an HTML form and apply the token So we get to skip all of that stuff and we don't have to enter the FQDN anymore So we've taken out a lot of the manual steps the entitlements is the only thing that's going to remain as A manual process because the node has to be registered with the controller before you can even get the entitlement So that's going to stay the way it is so To wrap it up. I just want to talk a little bit about Our release Dell Services Edge 1.2 It's it's based on the latest Version of Smart Edge It's you know zero trust Including TPM support and it's a full it's a mech as well as a full orchestration platform We you know most of the time in the lab. We're deploying right and we're provisioning But what workloads I've seen people do with the Smart Edge is pretty remarkable some of the Machine learning and inferencing and all these things. It's pretty mind-blowing some of the workloads. I've seen run on Smart Edge again, we're using OpenShift 4.8 for the controller host and We're also partnered with litmus edge for providing IOT VeloCloud SD-WAN is also an option and Airspan is our radio provider for this release We're doing 7.2 split and we're using CVRS in the US outside the US It's it's still the private spectrum is being talked about but There's no clear standard CVRS is a USA only standard and basically what CVRS is is a Free block of spectrum that you just need to register with what's called a SaaS service for your county And you're entitled to use CVRS in a certain band there for free This is a high-level topology of the design So on the right hand side we have the The network edge where it's on-prem and we have Red Hat OpenShift running the litmus manager Airspan ACP, which is Airspan's EMS And then the Smart Edge controller And then on top of that we have a VM running the ironic Bifrost The red arrow I have there is the the change that we did so this slide already existed The red arrow is is where we did new stuff to do this zero-touch provisioning The rest of the architecture if you were to go to the RA you would see this slide without the red arrow So the environment is using the latest version of the controller, which is 10.24 Layers version of Smart Edge Edge node, which is 5.6 Otherwise you will not get the features. I mentioned they get us at one step closer to zero-touch We're doing a simple single node controller But our RA the official release and our deployment guide we deploy the controller to OpenShift And as I mentioned we used XR2420s in our example, but the RA specifies support for XR11 and XR12 The actions that were taken To you know build this presentation was to leverage Ironic's boot from virtual media To activate the node via the REST API and you know as I said the remaining manual step is we Get the entitlements from Intel and then the node is ready to run That is all I have for slides Does anybody have any questions for myself for Rick on on Ironic or Akalash on Smart Edge Or about the new business unit at Dell No questions Okay, I hope it was interesting. Thank you very much Actually, all of this stuff here is all public. You don't need any kind of Credentials to get to it. Have a nice day. Thank you everyone