OpenStack Basics - An overview for the Absolute Beginner

all right thank you everybody for for coming to what I call this thing OpenStack basics it's for the absolute beginner when I say absolute beginner obviously if this is your first day NIT we're gonna this might be a little still a little difficult for difficult for you but some assumptions you know you have to kind of know what linux is or you know you have to have some basic knowledge but beyond that we're starting at the beginning so if this is your first summit it's not your first summit if you you know we're just given this to do this will be perfect for you this is a picture of me with a bunch of words I talk about me what I've done long story short build a cloud an American Express a very big one back in 2013 2014 and then from there I went off from Aransas to help them build their distribution and do a lot of architectures for fortune 100 and now I've written a few books and done some other things how I you know how I got started in OpenStack I think is important as I was at American Express coming off a 1 billion dollar project and was looking for something else to do and I had an executive sponsor come up to me and said hey your next project is going to be building an OpenStack cloud we're gonna be moving everything off of our legacy platform into this and you know it's time to get started so let's go luckily I had about I don't know three weeks until I started the next next phase had to start this OpenStack project so I thought no problem I'll do like everybody else does right we'll get on Google I'll have this thing licked in a week right I'll learn how to install OpenStack I mean they give you a CD don't they they just put it in and then you press go and it get installed and then there's a big manual that tells you how to do everything and we'd be cool so I got on Google and after three weeks I I look like that I was very surprised debating it back in 2013 or 2014 that yeah there were video there were videos I talked about ah slow messaging and how you know you know how to configure it and you know how to install OpenStack by hand and you know this projects the greatest and this project will do this and do that and has interoperability with this and that and I couldn't even I didn't know what a project was I had no idea shortly afterwards a couple months later I was sent off to one of the summits and when I got there as you know some of you have found out if you've gone to other summits that the beginners tracks are a little wanting in my opinion there's things for beginners but it's more walkthroughs on installs and things like that they don't talk about the what and the why and how did we even get to where we are so why cloud what did it start you know we're we just happy with our appliance based or our hardware based servers like we just we go get some HP s or Dells or something and we put them in racks and you know we hooked them up and we get operating systems on and then other guys come in and put web servers or app servers on them or database servers and then we we run our apps right and that's cool and if we want more we just order more servers and we put them in racks and I for some of you I'm sure getting to you smiling that's like you know that's really how we used to do it does believe it or not and it wasn't that long ago so you know we're talking like web 2.0 days where you know we have applications web scale applications running and then maybe we might hook into other api's for services to pull in WordPress or pull in other different third-party applications to work with our static web pages that's that's kind of web 2.0 but the problem with web 2.0 is this scalability is a problem efficiency was a problem you know how do you get these new servers you know RAC and what happens at Christmas or what happens if you have a product launch and there's a whole bunch of demand well for the most part wasn't a major problem because the way we did scaling back then was let's make sure our hardware based systems they run at ten percent of capacity during off-peak that's real efficient so 90 percent sitting idle and then during peak times we're gonna set up our monitoring's we hit 75% you better page me because that's scary oh no we're at 75% then we hit 75% once what do we do we got our provisioning system we order more servers but we have to get approvals first and then we have to find out budget and then they have to somebody has to build them and then they get them sent in and then they got put on a truck and then you put it on a handcart then someone screws them and they don't have enough screws you got to wait and then they cable them but there's not enough colored cables they switch them around 55 days later I have a new server meanwhile now all of my servers are at 75 percent or 99 percent or whatever it is and we've got a problem not efficient not scalable so and again as I said there's a bunch of approval workflows it could be sitting on people's desks for forever so by 2006 Amazon had this problem Amazon said this is ridiculous why are we doing this there's got to be a better way so what they did is they looked at their systems and they found out that again they were only using 10% of their of their capacity during off-peak and at most 40 to 60 percent during on peak that's a lot of wasted systems so they wanted an easy self-service way to provision what they had and to you know to get it into the hands of developers and users so that they didn't have to order new machines so they started AWS and AWS was an internal project so that they could have an internal cloud to build their you know retail services on a few years earlier there was a company named salesforce.com in 2000 or 1999 they created a service where they build it so that you could have software as a service so all the applications didn't reside on users servers they resided on on a web page where you can go or one a website you can go in and give yourself serve yourself applications so at that point they were you know they working on this solution and they couldn't they couldn't really figure out how to do it in the current construct so you know the they said earlier the challenges were were hardware based or or virtual virtualized environments and you know it was it was really a challenge for them to grow especially at the rate that someone like Amazon was growing so in 2006 as I said they came up with AWS they they were facing these kind of issues of scale because they were AWS most of the companies today or even back then they weren't feeling that type of scale problem they weren't growing as fast as Amazon so there was most other companies were laggards however eventually it caught up so we talked about cloud computing them starting a cloud and I think this is a very important part of the equation that the value of cloud computing is in the outcomes enables you know as I said if you're an app team in a company back in the the pre cloud days what would you do if you had an idea for a new app you would first tell your boss your boss would say great great idea nobody else has it let's get strategic advantage let's get you working on it okay I need something to work on and eat hardware okay hold on let me check my budget you check this budget okay I've got budget go ahead and order it when's it gonna arrive three weeks okay I'm gonna do my local card I have nowhere to test it all right so it comes in and they start testing it and on their local systems they get it all hooked up and then they need someone to install the OS that's another team and then then themselves or they hire someone else to install the web and app and DB parts of it by the time it's all done as I said we're two months into this guess where their strategic advantage went it went away about five weeks ago somebody else you know who had spare hardware got their idea first so the value of cloud computing is these the outcomes that enables and the outcomes it enables is it's kind of its kind of like similar the value of an elliptical trainer right so I like to call this as weight loss as a service and you know you you get on a elliptical trainer and so you know it's like what you plug in the elliptical trainer API and you do your workout and what comes out the other side is less weight calories burned same thing with a you know in cloud computing it avoids the problem of dev-team ticket requests into the infrastructure teams and the infrastructure deployment dev-team adding tools middleware application all of these things are possible for the self-service and you know we'll see the next slide here you know what what is cloud then so if cloud computing is a self-service platform where you can get infrastructure quickly and scale it rapidly whatwhat is this cloud we're talking about first we want to clear up what cloud is right so the best definition I heard and this was sort of inspired by a Red Hat definition was the clouds are pools of virtual resources like power you know memory network storage or applications orchestrated by management and automation software and there's four different types you'll hear that there's three different types but I'm arguing with that that there is not three different types there's actually four now since the advent of private and public cloud the first three came up very quickly public cloud it's a shared resource pay-as-you-go we know it as AWS for Google Cloud or Azure private cloud is a dedicated cloud to you where you get to control the resources you don't pay as you go but you have to make an initial capital investment hybrid cloud is a mix of private and public orchestrated together whether it's private on-premise or private posted and then multi cloud and again this is a very large distinction is use of multiple cloud computing services and a single heterogeneous architecture meaning hybrid cloud you have your workloads spread across two clouds but if you're using multiple clouds in your company to do an overall delivery of a service that would be multi cloud they don't have to share a workload so for example with hybrid cloud you may have your web servers in AWS and your app server and your database server on your private cloud servicing requests in from cloud front or something like that where multi cloud is you have your your web app in DB for an app of one sort in an AWS because you've decided it's closer to the you know closer to the consumer needs fast access you don't have a database there but another part of that app is another web app and database that runs in your private cloud and feeds data possibly into the multi cloud architecture so those are the kind of the differences that we see clouds and cloud technology they aren't technologies unto themselves they need operating systems hypervisors virtualization software automation and management tools operation officer operating systems they set up the network and host user interfaces virtualization UPS tracks the resources and then pulls them into clouds and then automations resources automation software allocate the resources and management tools provision new environments so I think it's it's interesting you ask somebody what is cloud and you will literally get like you ask 9 different people you get 9 different answers you have 9 different companies you get 9 different answers but some of them have common threads and just to prove this point I'm gonna read some definitions from some major companies in the OpenStack and cloud arena so you can see how everybody answers differently very shortly Red Hat says clouds are pools of virtual resources such as RAW processing power storage or cloud-based applications orchestrated by management and automation software so they can be accessed by users on-demand through self-service portals supported by automatic scaling and dynamic resource allocation whoo Thank You Red Hat for that that's a mouthful amazon says cloud computing is the on-demand delivery of compute power database storage applications and other IT resources through a cloud through a cloud services platform via the internet with goa piss pay you pay as you go pricing oh wait a minute Amazon added something you have to get to it through the internet and it has pay-as-you-go but read bet hadn't said no thing about that well what does Microsoft say cloud computing is the delivery of computing services servers storage databases networking software analytics and more over the Internet the cloud companies offering these computing services are called cloud providers and typically charge for cloud computing services based on usage salesforce says cloud computing is a kind of outsourcing computer programs using cloud computing users are able to access software and applications from wherever they need while is being hosted by an outside party in the cloud and Google finally says that cloud computing as a capital investment in building and maintaining data centers replaced by consuming IT resources as an elastic utility like service from a cloud cloud provider including storage computing network data processing and analytics and application development machine learning and even fully managed services ok so the common thread right they want to talk about data they want talk about memory storage hypervisor virtualization providing it getting to it by some means or another so the the the definition is very fairly unclear NIST the National Institute of Standards and Technology helped us out a little bit and set up kind of four I would say guidelines for what cloud computing is and they they they first call it the first day on demand self-service so if it doesn't have on-demand self-service if a user can't easily manage their own services without swivel chairing like filling out a form and handing it to someone else and saying hey you create me a VM not a cloud broad network access having access to virtualized or installed networks accessible via the cloud platform so if you have to call up a network group and say hey I need this VLAN run to my VM not a cloud resource pooling what we talked about that getting all the resources together and then giving slices about to to other people if you have to call someone and say I need more CPUs I need more physical CPUs that is I need more memory and my server yeah not cloud not cloud at all and then the fourth one is managing or having metering so measuring how much you're using and able to show you and either show back or chargeback model what you have so this is you know why cloud just in summary again efficiency scalability we can talk about cost a little later things like that we're driving this type of development what is cloud these are the four types of clouds we've talked about what is cloud and what is cloud computing what everybody thinks it is and what are the guidelines for it and then these are the the types of cloud services so when we talked about outcomes as a service right weight loss as a service on a treadmill or on an elliptical trainer you know you what you put in it you put your body on it it runs what do you get out of it you lose weight heart health whatever it is here's legacy you have a pointer here here it is here's legacy IT so here's what the that says customer managed but this is really what your IT group would manage right they manage everything they would install your applications for you they would do security to put the databases in operating systems virtually all of these things would be the responsibility of the IT team before they even gave it to you infrastructure as a service well now we've eliminated some of these so provider or this would be more the the IT group here would be providing this so they provide you with virtualization the hardware to run it on the stores Road on the networking to run on and they put it in a data center for you but now as a dev team or our customer of the IT group this is what you're responsible for so your responsibilities have shrunken you don't need it you don't need to know virtualization server storage network and data centers anymore just just this in is now platform as-a-service this is where you're delivered all of these things so a real life example would be if you wanted if you were a bit if you were a developer and you wanted a Red Hat Enterprise Linux 7.3 server running tomcat apache or some other lamp stack or something like that and you wanted it all configured and all you wanted to do is install the the jar file or or just install the the application or the nodejs application whatever it is this would be your this would be it here so then you would have access to just deploy applications and then software-as-a-service this is the salesforce.com the example where you basically go to a web website or you know this would actually be like the Google suite Google Google Drive Google slides Gmail even even office 365 this is a type of software as a service so these are the types of as a service well as a service we see now so I want to just make sure that everybody kind of understands what a cloud is what different types of services it provides and how they're how they're different each one of these topics we could go on for hours and hours and talk about but we're gonna kind of fly through them we have a lot to cover so here we go here's the meet right this is what you're all here for to learn about AWS I don't mean OpenStack everybody's like yeah no no not AWS what is OpenStack so this I like this little picture it's very very simple so OpenStack sits on top of programmable of a programmable network compute resources storage resources and developers and admins go in through the api's and in provision and develop on it the foundation says it lays a common set of api's on top of compute network and storage and it's one platform for virtual machines containers in bare metal so with OpenStack you can create virtual machines you can create containers or using one of the projects that we're not going to talk about today but it's called ironic and the reason it's called ironic is because we used to just do virtual machines and then they came out with bare metal and for a virtual machine platform it was ironic that we would provision bare metal but okay high-five for the name so but it can do all three of these now so again let's think back to web 2.0 days legacy days you want to create virtual machines what was the default the de facto standard right VMware right we would go to vmware we have VMs containers well once a containers didn't exist because that would be a misnomer but they were mostly living in sun so they were called containers there or virtual hosts or whatever you want to call them and then bare metal well that was that was legacy that's what we used to be doing right we were provisioning bare metal how do we do that we kickstart them we would jumpstart them we put CD ROMs and we would ghost them anybody remember ghost yeah it goes all right windows boxes windows 95 so but now OpenStack is one platform one orchestration platform that does it all from soup to nuts so why is it what OpenStack is open-source what why is that mattered right we here open source open source open source all the time we here open source gives you choice and control over your underlying infrastructure you can switch vendors you can add more vendors if you think today of hyper-converged offerings right you have these hyper-converged offerings they give you a box here you go here's your hyper coverage well what's in here never mind don't look under the hood you know is it so it's all HP don't worry or it's all it's all Dell or it's all Newt annex don't take the screws out you'll void your warranty so I mean that's the difference right appliances or hardware based now you can control it because you're laying an orchestration layer on top of it what runs underneath it it doesn't matter developers will never ask you again I can't promise that but you can always tell them to go away now when they say well what's this VM running on what does it matter you've got your V CPUs right but is it on Dell is on HP what does it matter yes yes what if you have a hybrid and it set up yes it's on all of that yes it's all very good thank you go away you know your ability to contribute never again if it's open do you have to say man I really wish I had this feature I really wish I you know I wish they had this button or this switch because if you can't find it and you can't influence someone else to write it guess what you can do an open stack write it yourself write it yourself get it in get it in your local you know get into your local development environment make sure it works I always suggest people contribute back to the community please please get involved and put a blueprint out for it and get it voted on and get it into the community but no longer are there feature requests that go into the trash widely adopted open-source api's rest api is i mean who doesn't use rest api anymore right there it's widely adopted everywhere and part of a vibrant community to share knowledge and help each other well that's why we're here today that's why I'm here that's why all the other speakers are here we're not paid we come we we do this out of our you know belief that this is a great platform and then we want to help others for OpenStack principles you'll hear some some about this if you go to other beginner sessions open source well that's kind of self-explanatory if you don't open sources see me after the lecture open design as I said if you come up with a great idea you submitted into the community or into the group or the project that you're working on we'll talk about projects later and it gets voted on and you vote yes and your brother votes yes and your cousin votes yes I'm not suggesting you do that no that's that's wrong but no the community votes yes and and it gets approved and guess what your patch gets accepted in your feature gets accepted in but there's no there's no real way to influence it because you're you know you're the cousin of somebody in the foundation does your feature go in first or get accepted over somebody else's it's fully open open development again you can go out and look at the source it's being updated on by the minute there are people out here if you look if you walk along some of these rows yes some people are doing their email some people are looking for shoes but there are other people who are contributing code as we're speaking here there are people doing patches there's people doing all sorts of things and it's updating the active OpenStack problems that come up they're fixing them as we speak right in here and then open community all of our meetings are held in IRC or in public spaces there's no closed-door meetings think sunshine laws think just fully open there's not a it's not a secret handshake for OpenStack at the Developers Conference anybody can go and you don't have to had to have a password to get in they let anybody in primary business drivers if you're if your boss asks you why you know why OpenStack right in the 2006 user survey that was released in April to thirds of users says cost was their number one business driver right I in at American Express I used to work that was the number one business driver they were being killed by licensing costs from their legacy vendor of virtualization I don't think I have to say the name but every feature every plugin every new license a new cost OpenStack cost model is it's completely different completely different it's open source you do what you want with it operational efficiency well we talked about that but spinning up new resources spending up new pooled resources right got a whole bunch of pool of resources you give a little here give a little there you keep some capacity and demand keep it off to the side hey it looks like reliable forecast this because the pool goes like this constantly people checking things in and checking it out need some development don't need it anymore it goes like this you're able to tell how much you need from plenty of time to order that hardware it get it put in and guess what there's no downtime you just bolt it onto the side keep bolting it on you can do that with you can do that with with compute and with storage and with network accelerate innovation again using infrastructure as code and again I don't have a session on that it's more of a DevOps topic but basically allowing developers to interact with these API services to create VMs as part of their deployment so for example I want to bring up a web server well before you had to get a machine but the OS on it log into the machine install Apache put the patch sets on do it do all this type of stuff but using OpenStack and using the api's you can have one one command that says do all of this put it all in make me a VM so when you're running a testing procedure like user acceptance testing that's the first line create my infrastructure run all the tests run all the load tests you know if it blows up crush the thing up put it away give me an alarm if it's successful show me the performance parameters then crush it up and put it away so it helps accelerate innovation nobody has to wait for infrastructure anymore you do it yourself you're given resources what runs on OpenStack I'm not gonna read all these out to you but everything everything runs on OpenStack mainframes don't run on OpenStack just to let you know yeah that's one thing that doesn't run on OpenStack but pretty much all the different use cases are here you can see some of these that use it some of the larger ones or Walmart and CERN over two hundred and thousands I think they're over 220,000 physical cores of OpenStack to 220000 so if you think of you know dual-core machines you figure out how many machines are running on OpenStack so if you ever hear anybody said OpenStack doesn't scale and wrong it scales scales just fine so these are some great user stories I suggest you go online to OpenStack org and look these up there there's some wonderful stories so where did OpenStack come from did it just spring up out of the desert did aliens deliver it to us as a package to fight the evil vendor Lockean aliens no it's close though NASA had it and they were working on a way of course to virtualize their infrastructure and they developed a product called nebula and it was basically the early version of the project called Nova and Nova is the project inside of OpenStack that creates vm's and we'll talk about that shortly but at one point in that Rackspace reached out to NASA and said hey we have this infrastructure provider service that we're working on the same thing would you like to hang out and have coffee and see if we can you know become friends and they said yes and then the friendship has lasted ever since and they developed OpenStack two years later the OpenStack foundation was established and pretty much it's been it's been history since there the marketplace opened up which is uh it helps potential users navigate what was it continues to be or like a rapidly growing ecosystem of vendors and products that run with around OpenStack and in OpenStack 2014 Juno was the first enterprise grade OpenStack that was released which people who had kilo running enterprise and enterprise will probably argue with that but that's the first one that was conceived as enterprise but even today you know half of the fortune 100 runs some sort of OpenStack is involved with OpenStack has it running it reduction or hazard in dev QA and today it's it's one of the premier platforms for private cloud for containers VMs and bare-metal so who are all these wackos that are working on this stuff coming to these conferences working on it well there's over 70,000 foundation members and 185 countries and now there's over 700 organizations businesses and organizations that provide support to OpenStack it's a very vibrant community and it's actually not even it's not getting smaller it's getting larger how do they release OpenStack you'll you'll kind of hear if you're new to OpenStack this will be a shock to you kind of but it's every six months a new version comes out don't panic nobody's saying you have to upgrade your OpenStack every six months you know they the OpenStack code is fully backwards compatible and anything that's not or anything that is will cause problems for upgrades is and at least n minus two so two two releases back right now where we are on coming up on Queens right now Pike is the release we're coming up on Queens will be released in February 2018 and in October it'll be rocky and saying well these are some strange names right Austin Beck's our cactus Diablo Essex Folsom grizzly Havana Ice House Juno kilo Liberty Mitaka Newton Okada pike Queens Rocky yes all the way from the beginning where do you get those names well they are associated towns or locations near where the summits are being held so apparently there's somewhere near here that's called Queens I don't I don't know yeah there you go thank you I knew a local would help me out I should know that but thank you so that's how that's done now I'm sorry oh yeah it's also alphabetical right so that's another challenge and being open if you're a member of the foundation they give you a list of like 9 trillion names and you're supposed to vote on them if you go back here Newton I did not vote for that I voted for null I would be an awesome name for for a version null like dev no like no like nothing that but then I think they thought well if you like run a command to find out what the version name was and it came back as null it might mess things up so good on them right the OpenStack framework so the people you know a common misunderstanding that OpenStack is a single product that it's uh Windows 95 or something you put in and you just install it like let me go get me some OpenStack or let me I would run down to my electronics store and get that box of OpenStack and I'll install it well that's not it's not it very modular and it's a group of nearly 60 open-source projects but we don't need them all it's completely extensible and modular so what you of the 60 that you need is based on what you're doing now we saw the use cases of what everyone's doing with it and each one of those use cases has an architectural baseline of the projects that they need some of those will be under sample configurations on OpenStack org but here we have the project navigator and this is on OpenStack org and it tells you what the name of the projects are and what they do and we'll be covering some of these very shortly but these are the some of the core projects and we'll kind of go over the naming but I just want to make it clear that OpenStack is not one thing it's not I'm gonna install OpenStack if someone says I have OpenStack install what that means to me is they at least have the core services which are six of and we'll talk about those but I have no idea what other services they're running I say well tell me what your architecture is tell me how its laid out tell me what it does and then I can kind of figure out how they're running it but you know the OpenStack mission is a platform that's fits every model everyone's use and you know as I said on the foundation website you can find sample configurations and they they have them by use case and they have them by functionality if you're familiar with AWS I know I've been picking on AWS people I love ATS by the way yeah taking my AWS test the end of the month for architect yeah a lot of fans in here all right you know tough crowd it must be Azure people here or Google Cloud right or Rackspace who's were actually tough crowd alright so if you know AWS these are some of the kind of there doesn't have complete feature parity you know ec2 and Nova but you can kind of understand them if you know the ones on the right you can understand the ones on the left and you notice I point out that that is not an actual AWS user anybody know who this guy is that's right for dummies right thank you aw I'm just trying to say that the AWS people are not dummies now Google and Microsoft might argue with me but OpenStack service overview so we're not talking about what runs on OpenStack that's a whole different discussion this is how OpenStack runs so there's what we call the control plane and then we have the workload plane or the user plane I've heard it's you know a lot of different ways for now is but this is how this is we're lifting back the covers we're opening the hood we're opening the kimono if you will and we're looking at what's under the you know what's under OpenStack dashboard right we all understand what a dashboard is it's horizon we login we got a website got a bunch of clicky things forms to fill in that's how we get our infrastructure that's how we make things happen and then these other ones we will we will talk about compute block storage networking image service object storage and identity service we'll talk about them a little more all right so anova we talked about nova a little bit nova was how to get VMs create them it basically provides the configuration for the VM so it takes all of the things that you put into the API all of the different parameters when you want a VM it takes all of that puts it into a configuration file or set of configuration files and into a database and then launches the instance that's this basis basic as I can get for Nova supports a lot of hypervisors and is a REST API service all of these will be REST API services cinder hell why they call it Nova I don't know it was called nebula and then it was NASA Nova Nova something in space terminology my biggest guess so when it comes to block storage look no further than your friend cinder here thinks cinder block this is how I remember it in the beginning cinder cinder block block storage cinder handles everything from the first API requests to the volume attach request to so attaching the block storage to the the VM that was created by Nova so this is that guy okay thank you Miss characters I had I mean I I had the same challenges as everyone here if you don't have these memorized or don't know what they are the names are not like one two or like block storage service like BSS or like VM service VMs or whatever you're just gonna have to memorize what their names are they won't change though they will all continue to say the same except for this one changed once before but we won't talk about that this used to be quantum but anyway Neutron this is the most complex one this is when I had the largest trouble with an in when I started it provides a software-defined networking piece neutrons the king of the networking stack formerly known as quantum provides Sdn functionality and instances and is extensible through layer to plugins and third-party add-ons I have no idea why it's called Neutron so anyone can answer that I'd love no but it always reminds me of like Jimmy Neutron you know I don't know if you have kids or anything like the guys always trying to invent new stuff like cooler ways to get places faster and route through here and all right that's the best I can do with the neutron you know he was very smart and inventing things and neutrons very slick very awesome when I started with OpenStack did not exist it was all Linux bridge it was all physical bridging he was called Nova Network it's deprecated now hopefully if you're getting started with OpenStack now you will never have to deal with it by the way Neutron does have that functionality it's called provider networks you can still go through Linux bridge glads image service you glance at images you you know when your google images you glance at them you don't look at them you glance at them so without glance there's no love for an instance because there's nothing to put on it Nova can not work without instances there's something you got to put on there whether it's rel Ubuntu Windows CentOS scientific UNIX you name it it can put it on there as long as it's in one of the supported formats and raw qqi so VHD VMDK etc etc Swift you know it's it's the object storage platform our project object source so it is the elder it is the OpenStack elder it's it's as old as Nova develop the store objects it's commonly used to store glance images glance image files it's fully compatible with AWS as s3 service if you ever hear anything talking about s3 or objects or object storage this is your guy in OpenStack fully distributed highly available as I said it's been around forever this is like oh man this is the daddy here's the gatekeeper this guy here without Keystone you're not going anywhere because you have to validate the Keystone before using any of the other services including horizon it is the it is it's the the authorization authentication engine for OpenStack natively and you pass through even if you're doing sam'l assertions or using LDAP or ADF all those things can be done through Keystone but what Keystone does and is most important is it generates tokens so for example you want to use a nova so say I want to boot a VM first thing it asks you is do you have a token no I don't have a token well you better go get a token user name/password here's your token well you know you have a token so we keep your token the token is stored in the database of OpenStack then you go over to cinder you say I want to create a volume token well hey I do here's my token so creates a thing for you two days later you come back and you say to Neutron I want to create a port or network do you have a token you say yes I do and it says sorry your tokens expired get a new token so there is there are X's default expiration is one day but you still need a token for everything that was a poorly delivered job all right simple view of how open sex services work they don't all exactly work like this but this is a pretty good representation of how they work in through the API you can do create read update delete of VM sin de Valle Yume's of storage let's just let's use the cinder example here with with volume the API communicates to a message queue puts it on the queue I need a new volume the scheduler takes from the RPC take takes via RPC from the message queue and says hey looks like somebody wants a new volume what volume types do we have out what storage we have out here oh they want you know regular storage they don't need high speed storage whatever is part of the request oh well I'm gonna put it on that net app over there okay puts it back out here's an NS brings it down to the service function says hey now I adds on hey I want this on the net app it's one gig puts it in here then at the service function it says okay I see that net app out there I'm gonna put it over here rights out to the database where it's going where it's gonna be and then pushes it through the driver for the net app which is configurable in the OpenStack configuration files and goes right into the provider store just created and then guess what all the way back up back through the driver to the API puts it on the message queue says hey I created it it's done here's here's all the information about it comes back it comes back as an answer through the API and if you know this you can if you can remember this that there's two stateful services meaning you know they need they need a static storage there static services message queue which you know that's call is stateful as well there are ways to make it not staple but we'll call it stateful the two stateful services then you have stateless services that are that run in this fashion you know how nineteen ninety percent of the the services or projects in OpenStack work that's it so you can follow it step by step you can even turn on debugging in the logs and follow the request step step step step so here's the architecture for OpenStack today oh no that's not that's Game of Thrones but if we could draw it this is what it would look like you have ironic over here at the capital and then the Tyrells they would be cinder and a nice watch but but that you will if you do a search on Google you will find something like this from cactus and Bex are and some of the olden days only when we had a few projects nobody has been brave enough to do all sixty projects and make one of these so that's why I laugh and say this is what it would look like there's Jon Snow right there alright real life example remember I said cinder sender volume creation I'm not going to go through this one through three four five but you can look and see you know user sender API messaging bus out of the messaging bus and there's cinder volume and cinder schedulers in there but cinder schedule or volume goes to the backend here it is right here's the three cinder volume four into the backend back in six seven back to the API nine back to the person all right it's not rocket science people like OpenStack is so confusing it's like so technical dude it's so hard to understand know if you understand this you can probably if you understand this and you know Python you can go start developing today so we talked about a few core projects here's the rest of them as I said I said you know I can do a whole lecture this would take me four hours to go through all of these so but people say well you know maybe somebody says OpenStack only does this well know it does all of this all the DNS service data processing you know Hadoop clustering service search workflow you know a lot of the stuff you see on AWS that is commonly used is here stuff you see on Google cloud is commonly used is here and then stuff that's not you know like the attacker that's an NFV orchestration service they don't have that in Google Cloud they don't have that integer these are these are things that people one day woke up maybe after a hangover or something and they said you know be a really good idea for a project did because I'm dealing with this at work I'm dealing working at Verizon AT&T Telstra reliance and we have all these vnfs that I can't keep I can't keep track of I don't know these virtual network functions how do I orchestrate them how do I get them the to go from one to the next or launch them in the right order well dude I can read it in pop it I can like bash script it out or you can write a project and get the thing done right and that's what people did and they just keep coming in so let you guys I'm gonna publish this deck afterwards it'll be attached to whatever wherever open socket a chah's at this time but you could take a picture every other oh all right so people say well all right I'm going to download OpenStack tomorrow I'm just gonna totally do all this and download it okay you can do that and Doc's taught OpenStack that org is your friend they have install guides for all different operating systems I'm a member of the docs team so I'm gonna give it a thumbs up but I will tell you that going through the docs and installing it by hand that's a one-time exercise maybe twice it's very very process and intensive you will learn a lot about OpenStack if you make a mistake you will learn even more about OpenStack but I recommend if your company's asking you to do something that you look into a distribution and these are the three distributions pretty much left today that are kind of OpenStack open-source distributions Red Hat su SE and Canonical's Ubuntu OpenStack if you're talking to a company or you have a company that says look we're a small medium business we don't have the resources to manage OpenStack we might have someone to install it we have developers to publish orbit but we don't have the operation staff we don't have anybody willing to take this on operations that said no way well first I would give their operations teams and raspberries and tell them no guts no glory but it's really not that different than managing Linux boxes however there is there are options and these are companies that provide managed OpenStack so they will come into your enterprise install OpenStack for you and manage it remotely so all you do is use it that's it you just use it there are differences between off all of these and I would suggest that you reach out to a company to help you kind of navigate those waters strategically onyx is one of those companies by the way not plugin but but these all have do they're all have nuances but that in general this is something that that isn't offered so I've made it and I've got some time for questions so this is where you raise your hand all right we have somebody with a microphone go ahead yell I'll repeat your question so I guess to you oh I know we went through this fast so feel free to ask any question you want thank you for saving me from yelling a short question so we have a small team I mean it's a medium-sized company but a small team that will be dedicated to OpenStack and we are small cloud provider mostly using the cloud and some other things and the question here is can the small team reasonably quickly onboard and manage this so we will start without the dev part and then later on get on that as well I think that's achievable absolutely you know the misnomer is that you need this huge team of Python developers and everything else to run OpenStack that's not true you just need somebody willing to learn the OpenStack architecture and learn the tools to manage it like anything else if you were putting VM ware in or V realized or anything you would still need people to know that as well so a couple guys that are dedicated to putting it in and running it could scale up to 10,000 instances easily not a problem I've seen it it's not a it's not a major task okay um I come from company who does radius DHCP for telco appointments between normal traditional boxes so now we have a bunch of telcos they're all coming back to us to to move everything to OpenStack environment okay so as we talk to all them or some say oh I'm open to do do to to audit Department some people are saying I'm gonna use ansible to do deployments so from a vendor point of view of a product how how do you actually sell into all these people because is that I will need to develop every single possible way of controlling my software to all the different configuration engines I think you're I think you're speaking about deploying the workloads no just to bring up the VM and how to interconnect the kind of thing because all them I starting let's say I run juju means I do do do deploy prayin stack right and then join relationships or use ansible we should run some play books but now I have my own code which will probably sit on either CentOS or open to base OS but I will need to deploy into a customer environment right and because I'm not running a single stack names I don't own the stack like some like far away or Ericsson or somebody owns and piloting and they have their own hundreds of solutions in their own stack so I'm gonna say in the open environments tank in these style coves so to bring our VM I add a do have a playbook to start up the brain right but you do deployment so which means as a vendor I need to develop every single way of know that's what the rest interfaces are for so the api's are what you would use to bring up your VMs in all of these different environments the REST API is the same no matter if you use Canonical's you know juju is just a interface the charms are just a way to line things up when you're doing workloads but that that's a choice you make but the API it interfaces with the API so there are puppet tools ansible tools chef tools there are all sorts of other tools that will allow you to interface with the API either graphically or through code to bring up workloads in in OpenStack environments the same to two summits ago they had a whole line of OpenStack foundation and vendors and everything and they told them on your clouds and on your software bring up these workloads go now and they all at the same time launched all these workloads and I forget who came out first second third whatever was it's irrelevant because of the hardware but the point is as far as interoperability they all came up and they all worked running the same commands running the same tools running you know all their own tools so there is no one way it's the way hidden you choose the way you want to do it I need my local environments in particular I need Mitaka version of it so can you tell what how closely dipstick follows the real application and because I really I will be all tasting a lot of is instable and deployments and all that stuff and now i5 flavor of image for that but I always need my local silo to play with right so if you've installed the Mint Mitaka version of devstack you've installed the basically what's out there in trunk which was out there on the wet OpenStack website the final version of Mitaka so if you're using that today then you're using the end of Mitaka if you go out there now today and download devstack you're gonna download the pike version that but that's the production version so dev stack is used by developers to code against if they're making printed project changes so it's as far as compatibility or as far as interoperability with what is really out there it's right on it's it's it's exactly what's out there I've been having a lot of troubles with particularly network installation scripts I think what you referred to as previous linux bridging i think it still tries to use a lot of nova networking as opposed to neutron for nova network to do a lot of I don't even remember now all the lines I changed in the installations yes Nova Network was deprecated in Mitaka so after Mitaka there was no more Nova Network you can't even you can't even run it's not there anymore it's you would use Neutron provider networks and if you look that up on Google it'll tell you how to how to switch from the Nova Network over into a rider networks it's actually much better it's not using the lineage bridging my customer they've got Mitaka okay well it's provider networks is available in Mitaka as well as nova network so either way but without knowing your problem like I can't dip stick just so did the dev stack that you like if you download devstack right off the site right now Mitaka devstack it's exactly the the Mitaka that was a geat yeah that is that was there was an end of life for each one of the recent of the releases so at the end of life of that release that's where the devstack is right now it never went any further and either did their version so that's it okay sure yeah okay cinder is for block storage at the end the question was cinder swift glance what's the difference between them they're all storage but what's the difference well cinder and Swifter storage cinder block storage swift is object storage and Glantz is the image service so it catalogs where all the images are for you know your Red Hat your CentOS whatever things that you're gonna load into a VM it can point to cinder or Swift so I'm sorry yeah cinder or Swift so they can be held an object storage or it can be held in block storage block storage block storage is storage that refers to that the the actual location on disk so databases file systems things like that they go and block storage the operating system sneaked operating systems they go on block storage object storage is for images video files HTML files actual file storage you can't put an OS on object storage you can't boot from object storage it holds things objects and not block locations that's probably the easiest way to describe it no no Swift and cinder are completely separate so Swift it's it's storage is on either Seth which is a distributed storage platform or can be on a net app it can be on an EMC VMAX the back-end storage is irrelevant the way that you refer to it is different so object storage objects have addresses that you would just refer to the object block storage has a block location on the on the storage device yeah they're independent yes they're independent yes yep like I'm sorry about the template is the template yes and no so you create images you create images that you can put in glance that have layouts in a certain way right you can have them have your security fixes or in your company maybe you have certain required software that has to be in an image you put that in there but creating a VM and put it attaching networks and things to it there's another service called heat and that's where you put in code to say launch the VM with this image attached this storage attached this network can figure it this way and then out the bottom of heat comes a fully configured it can even be used for platform-as-a-service install the web server install the app server and at the end you have exactly what you need but that's the that's the VM orchestration service where glance is just telling what the which image you would like to use I'm sorry the yes yep yes yep yep like cloud formation in AWS where you're putting together a bunch of different commands to get one output you know a VM or a workload configure it in one way that's heat there's another one called Murano which is similar but that's really geared toward a Paz platform and it allows you to drag and drop I want this image with Oracle and Tomcat and you know and then go and then it'll put it together for you that's another project that's belong it's in those 60 projects anyone else yeah yeah yes so as far as the message viewer and as any message in queue goes whe when you request certain services from the the different api's as you can imagine I have a thousand users on my OpenStack platform right at any one time 200 could be requesting volume creations deletes grow shrink whatever there is no internal buffering service on the services themselves so that what would happen is all of those requests would come in at the same time and they would be backed up so what the MDM the the queuing agent which is RabbitMQ what it does is it takes requests and then feeds them based on you know first-in first-out type of things like that you know the it holds it in queue so that it doesn't get backed up it allows proper flow it could be it could be even you know even with with rabbitmq in the early days there was something called cupid which was the early RabbitMQ if the timeouts were set too low the the requests will go in and before they were even done creating like a 50 terabyte it would timeout and then it would go back the volume would be created but it would send a message back and say nope didn't happen and then it would come the requests would come back again and say created again and then we start creating it again and then it would fail and then there are these problems so what it does is it provides a buffer between the api's and allows you to send status back and forth without you know having it real-time just never a good idea to have to not have any cushion in there that's why most applications use the queuing agents anyone else I would look at dockside openstack dot org first look at the install docks look at the operation stocks then what I would do is I would learn these core these core projects you've taken the first step obviously download devstack get yourself you know get yourself open stack on your laptop you can load all of OpenStack into one VM you can have it running you can even create instances it's gonna be slow as def I'll be able to be honest with you if you have it in a little VM on your laptop you can't expect much performance out of it but you can have a fully operating OpenStack or open stacks I've run five open stacks on my lap go in this MacBook here just to show people that you can run OpenStack multiple copies of it and and just start playing with it that'd be the first thing there are some great books out there OpenStack for architects is one I know cuz I wrote it but learning OpenStack is another one there's videos out there there's a there's a sweb site called learning OpenStack org I think it is that you can go to YouTube now is great better than when I was at the 2013 when there was nothing so there's a lot of a lot of great things out there also there are companies that provide training there companies like onyx that will come into beginners courses for you at your company you know if there's something that you need they do customized training as well so if you say well we need Sdn training we need a lesson on Neutron we need you know things like that there's always that option but yeah today the resources out there for learning OpenStack are fantastic on the on the internet anymore one more question I would have to say depending on their workloads so if their if their workloads are primarily what there's a week we call it cattle versus pets that's right that's something you'll hear this week a lot cattle versus pets you know I don't want to be insensitive and use cattle but let's say chickens versus pets you've got livestock so when when you create workloads they get created and they can horizontally scale like web you can put behind a load balancer and scale it horizontally and nobody knows the difference right one axe up you shoot it in the head it goes down you bring another one up nobody's the wiser if you have applications like WebSphere WebLogic things that are these types of persistent applications that know about each other and care about their partners those are not those are not cloud native type good apps for OpenStack now that being said OpenStack is plenty fine as a virtualization platform it has some of the tools now that VMware has and storage DRS VM DRS all of that stuff can be can be kind of engineered into it but it's more you say when would it be a time for them to change well one when they've run out of money paying VMware if there's any VM wearing people in there sorry but you know when they when they get done paying VMware all the high licensing fees or when they want to save some money taking some of those workloads that can can move over that'll be it that would be a good start for them and then these cloud native workloads that they can move over that are not pets they're not servers that if they go down you want to log into them and bring them back up they're not they're you know Apache lay up stack type stuff vnfs the virtualized network functions that they're making cloud native now these are all things you want to put on an open stack stack and get them off of a more expensive stack so migrate from VMware to into OpenStack VM okay right like okay okay Maya and something I'm sure what your might let your migrating you you have a public provider customer VM moving them from the public provider to your local that's easy I mean that would be I'm one say it's easy I mean it's everything in reason but you could basically make snapshots of their images that are running on this provider and then bring them over and put them on your private cloud in in the state it's called a nova snapshot you make a snapshot of it and bring it over now there are ways to do it a little bit more gracefully there's a project out there called cloud ferry that allows you to go from OpenStack to OpenStack but there's I mean there's no way alive migrate them like you know move them from one cloud to the other but the clouds Ferry would be the closest thing but if you can take some outage you know create an image and move it over shut it down you know replicate the storage bring it up it's it's as easy as that it's that's from OpenStack to OpenStack is easy even VMware to OpenStack is easy well thank you everyo so we have another question I'll take one more the discussant who does the yeah seems like it doesn't it Rhett Red Hat doesn't have a certification right I'm sorry red the red hat does have a certification Moran tiss used to have a certification I think they still do and the foundation has a certification the certified OpenStack operate administrator certification those are the three that I'm aware of now there are some other ones out there that are not as popular but I would have to say that a popularity red hat first Maran tiss and then the COA exam from the foundation is kind of a no-brainer you want to get that because that's actually from the foundation itself and the COA exam was put together by all of the different distributions they contributed to it to make this one exam so it shows proficiency all right all right well thank you guys for staying around and thank you thank you thank you

Channel: Open Infrastructure Foundation

Views: 74,549

Keywords: OpenStack, OpenInfra, Open Infrastrucure, Open Source, Demo

Id: 8kADjGCuSVI

Length: 71min 57sec (4317 seconds)

Published: Sun Nov 05 2017