Drut Webinar Drut Software Platform Install
Good afternoon and welcome my name is Simon McCormack I'm product management for Drut Technologies and today we're going to talk about the Drut software platform the DSP as we call it internally. DSP is a really interesting bit of software that you can use for building private Cloud type infrastructure and it's the base software we use for Druts photonic Fabric and disaggregated solution. What's interesting is you can run it as a standalone entity so it doesn't always have to come with the photonic fabric it can be a building block towards that. In the webinar today I'm going to basically show an installation of the DSP and show how smoothly we can deploy this platform and the different components are part of it, it's kind of an interesting thing to show a deployment but we want to show how the bar to getting a private Cloud setup can be very low with a DSP type environment and it really help you towards building your own private Cloud environments. Before I go into a demo I want to do a few slides to tell you a little bit more about some of the components of the DSP you know because there's a certain amount of Drut technology plus some open source technology so if you're going down that direction for the open source then this could be the platform for you.
I'm working product management so I spend a lot of time talking to customers about their problems and trying to build products so we meet the problems that these customers have. So I wanted to start with a little bit about about what's driving the problems that we see, what's driving us building this software because you'll see then you will know what the direction we're going and why we go in that direction, if you think about some of the problems that we're going to describe. So I start off with this first slide, what are customers actually want from a private Cloud type deployment. They're all looking and you know you can say some have different problems they all have different problems Etc I say they're all looking for essentially a cloud-like experience so they want their users application deployment teams to have a simple interface to deploy the machines that they're going to use for their service we also think we can extend that kind of concept to all layers in your organization so you think about people are involved in your data centers they'll be low level infrastructure people people deploying Cloud platform in addition to the application deployers and the the users Etc. We think they all deserve kind of a cloud-like experience something simple for them to be able to deploy what they need without worrying about the underlying infrastructure that's a lot of what was driving people towards I think the public Cloud because that gave essentially a simple interface to get into it so that's a one-core driving Factor behind this software is to try to give you that better experience. There's a number of other things customers want you know, the list is endless but I've got a few here there a lot of people are very frustrated when they're purchasing and they feel they're locked into Technology Solutions so more and more customers we're finding they want more standards-based flexibility in the software Stacks that they're deploying which is where open source becomes interesting because we've seen a lot of maturity in the open source software throughout the industry in in the last few years and I'm going to demonstrate some of that as I deploy the software today obviously they still have high reliability and performance needs ha is still high none of that changed they want control of costs so one of the problems we've seen is customers getting unexpected bills in in the days that they are today and they also want a mix of public and private type services so that in most cases there's no concept that you're all public or you're all private you wanna they want a little bit of a mix of the two and security is very important to them so there's some I've got six kind of big bucket items about requirements here and the big question is why does public Cloud not satisfy on its own and what are you what can you do with private Cloud to meet some of these requirements so my next slide I talk a little bit about some of these needs in the public Cloud space what we found is that you know it's continuing to be more expensive than people thought and they're getting unexpected bills which are ever increasing and they find themselves essentially locked in once you've made that effort to move something to the public Cloud it's very difficult to bring it back and as you make it some kind of hybrid people do have security concerns obviously if they've got their own private data you don't necessarily want to put it in the public cloud and some would say just won't mind public Cloud it's just too much too big too difficult and there's so the result is in general whatever happens you have to maintain your own private clouds going forward I think there's a lot of Industry talk over the last number of years particularly the analysts talking about everything moving to the public cloud and obviously that never made sense I think the private cloud data center industry was continuing to increase quicker than the public Cloud well but all the Press was about public Cloud because obviously that was starting from a very small base I think ultimately you'll you'll always see both together but you have to manually maintain your own private Cloud to keep your business going up but I mean people still have problems with the private Cloud especially today and the climate today it's you know private cloud is not cheap there's a lot of costs there Power Cooling particularly with the new GPU type loads that are ever increasing number of power ever increasing need for cooling that equipment and not just withstanding the equipment that you need to purchase for these environments as well so it's all very expensive very complicated because there's lots of components from lots of vendors you can put together and not everybody has the Personnel to to do this either I mean a lot of the times people have cut back on their resources so it's very challenging and I think we think there's a change in the the traditional monopolithic data center for Enterprises so the hyperscale has got these massive bigger and bigger data centers we know that but for you as an Enterprise the type of data centers you have today may be different your traditional data center they used 10 years ago may not have the power and cooling for modern equipment making putting it all in one place may not even work so a Colo space distributed data center makes a little bit more sense so you may have little pockets of data centers all connected with high-speed connectivity to make it all work and I and I think this This lends itself well to the Druid solution because I want to be able to then deploy easily a new little private cloud and that's what I'll show you in this demo today when I go through the product enough about requirements what's Driven us here I think it's the the primary thing message that I wanted to give there was basically you know there's a continual need for private clouds in private data centers they may be more distributed so you need small little pockets of connectivity and Drew actually has software that I think we can help you help you with this kind of situation so the bottom line number one thing is that we've got software that we can build for you we can manage it for you but we want to leave you in control so you're using a lot of open source gives the control to you because that open source you continue to take forward so what we've done at Drut is because we've you know we've built why we built this software stack in the first place is an interesting thing to consider with the Drut hardware we had to build enough software to be able to use our Hardware right the base level it's an infrastructure type software that we built anyway and what we found is just that infrastructure software wasn't enough to validate our use case so we built more software on top of it more layers of software to prove an application works in a droot hardware environment so we built the software anyway for our own purposes we showed that to customers customers purchase the software and we realize we need to productize this solution because there's a lot of value to what we're providing so I'll talk about the different layers of stacks of software that we have in the solution my charts here I'm trying to describe in the middle there you see Drut software platform that's really the underpinning software stack that we can use to bring all this together and then below that fabric manager which is essentially optional and then on top of its services I'll talk about why that's there in a moment start off with the Drut software platform we basically have a hook into work or infrastructure and Hardware so if you've got data center machines you need to manipulate them at a hardware infrastructure level bare metal type level to be able to manage these machines give them addresses Etc so it's a bare metal type structure you need software to do that the DSP can do that for you so you can purchase this solution in these different layers so you could just use the DSP to do two basic infrastructure of your compute environment with lends yourself in the you should be able to add group fabric as well but you can continue on just using the root software the next layer we optionally put on is a that platform layer hypervisor containers virtual storage so we've got the Drut software platform experience where we can simply deploy this containerized or virtual machine or virtual storage solution together which is what I'm going to show as part of this demo in fact I'm showing one and two essentially today the top bottom two layers now the next layers Cloud lock experience have some workloads become very customizable per install that's where Services come into it I think everybody every customer dreams of a cloud-like experience their services what they don't really realize up front it's very difficult to do that for a vendor in a generic fashion it needs to be customized to you so you need services to do that to really pin it all together we can show examples and base structures but you need something customized same for apps and workloads which is why we have that layer built on top of it all so the components we're typically deploying in the DSP so there's a prime layer infrastructure layer to DSP which is typically fixed towards providing pixie Services image Services DHCP DNS networking power management for devices and also moves into the desegregated GPU type environment but then we have options on top of it if you want to do Cloud platform software on top with us then we can take you down this journey of doing the open source kubernetes openstack type environment you can choose different ones and you can do it yourself so I mean we can give you a base platform that's basically the hardware uh deployed and networking figured out you can then start deploying your own software on it or you can let us do the next level which is the openstack and kubernetes piece and then when you're doing something like openstack you've got to have a storage solution that kind of go together you don't really have one without the other at this point so Ceph is the storage solution of choice for us although we can be flexible the model we've built allows plug-in modules to be able to be inserted and you can do different Technologies depending on what you want to do on the left hand side you see networking because we have a certain amount of networking that will build into the solution a basic VLAN configuration subnet configuration and I'll show you the little bit of this in the demo the right hand side the cloud experience is obviously you as I said the services would tend to do that and then workload application services so my goal today in this demo it's primarily an installation piece I want to show you how we built a Playbook solution to be able to deploy the different applications once they're up I do a little troubleshooting session I wanted them to show how to get it up and running smoothly and as I made some let's call them intentional mistakes in my deployment to show how to troubleshoot the deployment mistake we're going to cover more details on these applications in future sessions so I'm going to do a deep dive on Ceph do a deep dive on openstack we can do the cloud experience portal and the other pieces I think as we go on so zooming into the demo wise and I what I've done this time is because I'm doing deployment it takes more than a couple of minutes sometimes to do deployment of the software and we don't want to all sit here where I did it live so I did it in advance and videoed it and I can play my little videos but I want to start off with talking about you know what happens in a data center type deployment typically most data centers are already there they've already got this function called management control which is the DHCP PXE boot image Services already in the data center they already got that solution this solution here comes with its own management control solution so how does that work well we want to be able to mix the two together nicely so they can play well together and still be able to offer these different Services now if you listen remember what I said earlier on in the presentation a lot of new data center deployments are little remote colos what's interesting the original colos because I've done some of these you turn up the cage the cage has a rack there's no server in there for you to deploy your management Control software on Twitter there is an existing function so you have to basically sometimes bring the management Control software and put it together so you almost have like a check in an egg situation where I got to deploy my management Control software but I need something to deploy it but there's nothing there to deploy it with in this case you can use a USB stick pre-install it on a machine or in some cases we use a laptop to deploy the first machine the da the demo here actually I did a video of me deploying the basic machine just gave me a Linux machine on a bare metal that's all I did so I I'm not going to even play that video because it just imagine I start this whole thing I've got one Linux machine on bare metal and I'm going to use that to bootstrap this opens environment first picture I'll go on to is actually showing it's a simplified view I've cut a lot of it out to try and simplify it so I've got my control management software which is my DSP that's just a Linux machine at the moment and it's connected to my management switch not worrying about storage or data networks at the moment because we're just trying to get it deployed we can talk about all that in another session as well I think that's probably worthy of an hour just talking about the networking pieces to plug all this together so here I've got a Linux machine and I want to deploy the DSP so I'm going to go into my video or before I do my video I want to talk about the broadcast domain so you'll see I've got a the way we set this one up is I've got my yellow broadcast domain and my kind of Bluey green broadcast domain I got two separate domains because I imagine I'm plugging this into an existing data center they've already got management control functions so I basically have that on a separate broadcast domain to my new function because I want all my new machines to boot up this dsb so when they come up they'll go pixie boot anybody out there and they'll hit my DSP they won't hit the Legacy set up now in my lab I have hundreds of these management controlled devices they're all over the place which is why we we separate it like this you don't necessarily drop hundreds in your places let's go to the video and we can see what he's doing and I and I speed some of it up so basically I've gone in I'm actually running an ansible Playbook that it's the configuration and manipulation installation of these machines so it's basically it's I'm on this Linux machine I've deployed I've downloaded a file that Drut gave me and I run this Playbook throughout the rest of the session today I don't need to manually download any more files it's going to grab everything for me so it makes it really easy I just grab one file put it on this Linux machine and then run it and it's now doing the management control installation and setting me up for the rest of it I'm gonna be able to set an openstack from this system if you've ever done sap and openstack you're trying to grab packages from all over the place installing different things figuring out the versions there's a lot of complexity behind it I I like the way they've done the Open Source Products and the so many modules you see this install is finished which is great I like the way they've done modularized it but there can be a lot of different components because not everybody needs the same things so we've curated this and put it all together so this is me logging into our DSP product for the first time now so I logged in as droot and it's saying DHCP is not enabled welcome to the solution I'm wanting to answer a bunch of questions in the the basic splash screen and then I can go into some of the networking setups and do things here your noted says a bunch of networks already set up so what it does is actually read from the Linux machine so the vlans and networks that were set on the machine it pulled them in and basically configured it I think there this solution that we have the Playbook solution is its first version we're going to continue to improve it one thing I did is I went on this machine and created these networks and then plumbed it and then it read they read them in here automatically what I'd like is to be able to go in here and add it and it push you to the machine make it go the other way around which is definitely within the bounds of possibilities on this and there's a bunch of other little things that you do but I've basically set up my uh my system that can do the deployments now so I can move to the next setup so in this world I've got to do openstack and I'm going to do Seth I want to set first get the storage layer built then I'll do openstack and then it'll integrate the two together running from these playbooks and I don't install any more software it's going to do it all for me grab whatever it needs got all the packages there anything it updates it will grab automatically so this setup now if you see I've created I've got these three new machines My Demo DSP five six and seven they're connected to three of the the yellow ports so the graphic is a little bit funny but there trust me they are connected to those yellow ports so that those machines when they turn on can be detected by the DSP so they'll beat they'll Boot and they'll do a PXE boot and it'll detect them now if you've got a server that's not set for pixie boot you need to set it it's a little pixie boot out of this port changing the BIOS story but once you set it so it picks a good out of the report it'll be captured and it'll be controlled in this environment you didn't even install an OS on these machines the pure bear mode you got straight from the factory let's watch the install left hand side you'll see actually I've got here's my machines on the right hand side I'm running my CEPH Playbook so the set Playbook is going to basically start grabbing the machines now in this case I told CEPH which machines to grab five six and seven I told it I said well maybe I want for the cloud like experience I don't want to tell you which machines to get this layer we're showing you is layer below cloud like experience Cloud experience as a user says just give me the machines I Don't Care What machines and that that layer would program this layer to do the actual installation so here I say deterministically do these machines so it's grab these three machines you've seen on the left powered them on once they come on it's going to install the ceph software and then it's once it's installed the software and got the packages right it's going to do the configuration and configure the ceph cluster together it's right now it's waiting see it's it's going a little bit black screen and I sped some of this up this is not not doing this in real time because really been waiting a long time but it's waiting for these machines to come up uh to be active and then it'll start doing the next session so the Playbook is pretty smart it knows what's out there connect to things it can wait based on looking for a status if it if there's a problem or an error it gives sensible errors I've had situations where I told it the wrong machine like I said dsp-8 and there isnt an eight it just came back saying you don't have an eight um and then you have to go and fix it and you can just rerun the The Playbook so it's found the machine they're all in the right State now so they're all in the right now so now it's installing the software so all I did to get to this point is I answered some questions basically to prompt it because I had to tell at the machines and a couple more questions were answered so from the user perspective it's pretty simple answer a few questions run the script and it does the deployment for you really removing a lot of the headache for getting these kind of services installed uh the way they were before so step is there and up and running and I know we've already had this question in in one of the in the Q and A's that we can see so I think some clever person said well how do you know it's deployed right well and actually in this case I didn't deploy it right I made a mistake you do need to go in and have a look and run through validation steps to make sure it's correct um we can do we've done a certain amount of that automatically but I've done something that only ceph could tell you so actually I didn't mean to do that I go actually to click on the next slide still didn't mean to do that idea there we go I found it I go to the next slide is it working so let's run this one because this will take us through an interesting little ride about whether this is working this is me logging in for the first time and I'll prove it because it makes me change my password uh when I log in I'm going to try and figure out you know what's the health status and how can we go about getting the help the status of the system uh to be uh healthy and everything working correctly so I just wonder I'm thinking out loud is that our installation script can do a series of automatic checks at the end of it and I think that would probably be a good thing just wait there and then come it comes back and tells you but in this case today I logged into the dashboard you see the dashboard's giving me a warning if you see the charts in the bottle it's talking about the PG status it's kind of inside baseball stuff a little bit about Ceph um basically you want to clean up the the placement groups these are actually it says placement group positioning and too few osds so the osds are the disks are on the setup I literally gave this set up two disks, I got three servers I gave it two disks and the minimum it needs is three disks normally you'd have dozens of disks you can have lots of service lots of disks not so maybe I wanted to see if I could break it a little bit and it's also an interesting exercise to see about fixing it now when I did this when I did this over I I had a bunch of you see it's only used two systems not three as well so I've got three servers I only use two for disks so I went about actually the first thing I do is add more disks so I went to my service and then this screen basically it's showing us OSD one six and seven is OST one this is the disk thing we're going to make that for two later because I'm going to go to the system add more disks and this one is complaining about the PG status it's undersized pgs placement groups which is where it puts all the data across this F data store so in in general I don't care about that stuff so now here I've gone to my machine I've added some disk and you see it's got a third disk there actually if I pause it a second you see this this third disk has been added in already state so obviously adding disks is very dependent on the uh the hardware you have whether you can hot insert them or what these doubt particular Dells I was using uh hot insert them but I but I had to uh this machine I have to reboot it and create a virtual Virtual Drive on the system and so having done that and rebooted it jumped forward a little bit it's basically uh added the new OSDs automatically the problem was that actually I I still had a replication factor is too high said replication factor is three I've only got two machines that have got OSD so I actually changed it all to two so that was my next step so my I didn't have enough thought placement groups or OSDs and I had the replication factor too much now you see I cleaned it up and it's all gone it's all gone clean so replication Factor two two copies of everything PG status is now all clean go back to the dashboard and everything's good so that was a quick method that I went through to to try and troubleshoot it I think what I want to do for surface I'll do a deep dive on this where we go through this in intricate detail because there's a lot of interesting use cases you can go through with uh with your data and volumes and machines and the high availability of the whole environment etc etc that is very interesting I just wanted to give you a feel so you think about from that that install where I did the the video where I did the install for the Playbook that got me here all I did was answer a couple of questions and it took me through the install when I got to the next step so it's very easy to get into that um position from the software and I only downloaded the one package from Drut to Kickstart the whole deployment and I know I was slightly less on my discs but I got that installed up and running now the next step I want to do openstack so if you notice I've got three more machines in my picture three open stack machines now so I'll do the Playbook deployment of openstack and then it's going to do an integration with Ceph bring the two together at the end of that I'll be able to create a VM in openstack using the data from a volume from the ceph setup so I've basically got a fully functioning setup at that point and all I'm going to do is run another Playbook so here I start off by actually looking at these machines so these machines so I looked at the machines I'm going to do for openstack I needed to tweak some networking for them because I want to do a little bit of network change to give the VMS access to the internet because we've sort of got external facing VMS now I'm doing them in the DSP UI this is before machines are even turned on so I'm kind of prepping it which is a nice function I I then pointed my questions that I did on the for the Playbook I then configured locally um and the two are synced up together I think an enhancer here is I can make one do the other so if I do this it should push that answer that question on the system for the installation and vice versa that's kind of the direction I'd want it to go now I run my setup openstack uh Playbook and it'll go off and do the same thing as as it did before similar thing it's going to grab the machines on the left powers them on deploys Ubuntu in this case that's deploying the OS from scratch all I did was give it some bare motor machines deploy Ubuntu and wait for it to come back it's it's happy with the machines come back I was really quick on the fast forwarding this time and then it's doing the next Playbook so it's basically got a series of playbooks here and the next Playbook is basically installing all of the openstack components and there's a with open set there's a lot of different components you can choose continue to add and change them as you go forward so there's a lot more pieces that it adds to the configuration once it's done that configuration then it completes it by doing an integration so making the volume from Ceph available Within openstack so that as a user I can create a machine and from that machine I want to give it storage it's going to use storage from the shared Self Storage environment I'm really making it simple for the openstack operator and giving you the basis providing that cloud-like experience to the user so they can do so they can create the machines and starting sorting applications as the one they want so you see there's a lot of tasks that's running through in this particular setup to get all this configured and put together so really as a huge Time Saver for an operator to be able to do this now we have additional playbooks to do more functions so the next thing you say what if I want to add another server tomorrow I actually got a playbook for adding a host and I think I'll show that again in a another session because those this first time installs are good this is a good basis to show you can get up and running sometimes you can get away with more complicated installs sometimes they'll give you a week to do the install if you're doing that in a Data Center and it's finished it's set up now sometimes I'll give you a week to do the install so you've got time to get it up and running once you're up and running in and in production or close to production those luxuries are gone you don't have a week to do anything now now you've got short windows so those are the operations that are actually more interesting than the base installed but I wanted to do this one first because I want to show you how how easy it is to get them running with the DSP now I've got this software base that I can start building on top of and doing a lot more with if I click the screen like that always takes me to back to the the back to the video again so in summary I think what we want to describe off for for droot is uh we're really more than just a hardware component company we've got a whole software solution private Cloud stagger software which is all optional pieces that you can build in to offering cloud services for your environment depending on your needs we probably got the software and we definitely have a lot of the hardware for you to be able to build a modern private Cloud that sets you up for the future and in a lot of cases uping open source standard type software that really puts you in control and not so vendor locked into other Solutions well that's all I had time for today um thanks very much for joining um I will see you again in a future session goodbye
2023-09-25 04:14
