hello everybody a warm welcome from berlin today we want to talk about it's dana trogonomic and what it has to do together with the cloud pack for watson air hops but before we go into the topics i'd like to have a short introduction round my name is fred klein i'm located in berlin i'm technical lead of the swat team and today i have three colleagues with me and i'd like to hand over to henning to introduce himself thank you friends my name is already listed here on the slide i'm located in the western part of germany on the other side of germany so to speak from from berlin's perspective working in the dark market for the so-called enterprise containers team so everything around container orchestration plus also the operations perspective with what's in there ops is within my technical view so over to you frank thanks henning my name is frank rickard i'm a sales engineer with turbonomic i'm based in frankfurt and i'm covering the dark market so and i'm heading over to peter hello everyone thanks for having me my name is peter yunkin i'm working as a solution architect at istana and i'm heading the technical sales team here at instagram for emea and yeah great to meet you and great to have the ability to talk about install so then i suggest to jump directly into a demo of installment let me quickly share my screen let's start with a quick installment um introduction so instantan is an observability platform and the idea that i have brought for today is i have here the kubernetes cluster and this is a three node kubernetes cluster and on this kubernetes cluster i have deployed an application a demo application it's called robotshop and let me quickly show you what's been deployed here here you can see the basic kubernetes system so i've just started this seven minutes ago and then a few minutes later i've deployed robot shop here into this robot shop namespace this is currently what's running and now the challenge is to have observability in the setup and within stana and here i've just started this watch demo environment it's pretty easy to do this when you log into instagram and you have nothing that is currently monitored instantan automatically suggests you the different options that we're having for deploying installer to this cluster so here we have all these hyperscalers we have for example vanilla kubernetes we have different other operating systems and whatever you pick for example in this case kubernetes we come with a variety of options how you can deploy in stanna into that environment so in this case for example for kubernetes i can just run this helm chart command it's a one-liner so it's not that interesting so let's quickly take the yaml file and give the cluster where we want to deploy in stannah just a name let's call that the swot demo as a name what i can do i can download now a yama file for this configuration so let's quickly download this let's put this into my downloads folder and so it already exists so let's override this there we go there is my configuration file and the only thing that i need to do to deploy in standard to my kubernetes cluster is i just run clip control apply i specify a file name it's in my downloads folder and here it's called configuration and that's it so what i've done now this pre-configured configuration yaml file that i've just downloaded is now applied to my kubernetes cluster and you can see what is being created we create a namespace for installer we create service accounts we create configurations and so on and that's all you that you need to do when i run this cute ctl get pause or name spaces again what you will see is you will see we still have our cube system we still have a robot shop and in addition to this we have installer agents these are installer agents have been deployed they are running so since 21 seconds here and now the instagram agents are updating themselves they are starting to download specific sensor depending on the technology and they will automatically begin to collect information going back to instagram what we can do we can click on this installer agent yaml file and then you can see what really is specified in this so here all the configuration is set you have a cluster name and you have a configuration yaml file which is mounted as config map so here when you need to apply specific settings you can do this by the challenge or the the approach that instant is following we try to do whatever is possible we try to do it automatically so you really should only it should only be required to make minor adjustments to this configuration form so as mentioned earlier now what happens the instagram agents are starting the instant agents are updating themselves and the insomnia agents start downloading specific sensors and here when i click to go to installer i can avoid having this specific configuration screen and what instagram will monitor it will monitor the infrastructure we will see this in a moment so we will monitor hose and all the components running on this hose and standard will monitor the kubernetes cluster and the platform as well as the applications so in stana will wherever possible automatically enable tracing and you can see all the individual traces and it's done let's start with kubernetes so when i click on kubernetes currently you see okay nothing is here so we need to give it another i don't know one minute okay now let's take a look at all the different areas in instagna first infrastructure what you can see here instagna has up to now detected two nodes and two nodes they are running specific or individual docker containers here on the left hand side you can see some changes so what instanter does in stunner starts instantaneous automatically detecting all the technologies that are used on this specific component and here you can see that these generic docker containers they change to a higher order technology like a node. js application like a redis like a revit mq and so on and this process is done automatically by a spanner initially and continuously which means whenever something changes instanta is aware of these changes currently we have two nodes the third node is currently the installer agent is starting up so we will see this in a moment this is the infrastructure in addition to this what you can see on the left hand side you can see the kubernetes area so the instagna agents have detected they are running in kubernetes obviously so one of the inspira agents is pulling information from the kubernetes masters in this case we saw we have specified a name what's called swap demo you see some kpis on of this i can click in this then you see the details here the metrics start from the point in time when we have deployed installer so we have plenty of information regarding this kubernetes in this specific case ga gke cluster and here you can drill down for example to all the pots that are being deployed on the system here for example we have the robot shop port i can open this port i can drill deeper into this and i can see specific details this goes to a detail level where we even have the spec and this is the spec for this specific kubernetes plot in instagram we are always showing you context so in this case you know this specific kubernetes port is running as or in this deployment in this namespace in this specific cluster and when you click on this kubernetes icon here you can see all the clusters they have deployed in style this is the infrastructure this is kubernetes and in addition to this we are providing application monitoring which is tracing so here when i click on these applications view you can specify your own views we call this application perspective so okay you can really specify very granular way which kind of calls which kind of infrastructure you want to see the application perspective that we are shipping out of the box is all services it contains five services and when i click on this all services you can see now relevant kpis that specify service quality so in this case very simplistic demo application in the last 10 minutes we had 359 cores 16 of them were erroneous which is 4. 45 and the mean latency of nine milliseconds and i don't want to click through all of these details but we have plenty of metrics and this is something that stunner automatically creates it's called the dynamic dependency graph so instagna understands what service is talking to what other servers and instagna understands dependencies by this you see some particles flying here the particles indicate how frequently service is called and if there are some red particles you know that there's a specific error rate and the point or the origin where we get this information from istana is tracing 100 of all the inbound and outbound calls so when somebody has used the service here that we are tracing we have a distributed trace for this so what i can do now i can go to analyze calls when i click on analyze calls we jump into instant analytics indicated here by this magnifying glass and sunnah knows we are coming from this all services application perspective and summer automatically groups this in a logical way so in this case it says it makes sense to group them by services and here in this case for example i can just remove the grouping and then here on the bottom we can see each and every individual trace and when i order them by latency for example here we have a request it took one millisecond at this point in time with the get request to this url i can open this and then we see a distributed trace what was going on for this in this case here there was an http request and this has made a get request from redis cache so this is what it's done provides all of the box the idea is to automate wherever possible to have a very high resolution of the data so all the metrics that instantaneous having let me open the dashboard again so hopefully you can see this here the metrics are moving uh this data here entering in the system of five seconds old the very recent data and the resolution of these metrics is one second very high data granularity we trace 100 of all the traces the next thing that i want to talk about is context let me pick an application for example here i want to take a look at a specific node.
js application this catalog application when i open the dashboard you can see here the metrics are moving the data entering here on the right hand side is five seconds old and the resolution of this data is one second so very high granularity of data we have one hundred percent of all the traces meaning inbound and inbound calls this is super relevant because once you have an issue you want to make sure that you have the data and installers providing context for all of these views having said this now we have deployed instantan to this gke cluster so now we have full observability into the systems and of course this observability can be used by several different other systems now i will head over to flag thank you peter so now it's up to on the point to talk about turbonomic and how to dynamic uh fits into the play so i have a small agenda for you i have only a one slight introduction to drogonomic to highly explain what problems we are solving and then i step immediately into a live demo because what you can see live is always better than a presentation and slides everybody can paint nice slides with that i'm going on with my one slide and talk about the common problem we see in the industry the problem is that everything is taking care about the application the application is the most important part in our world and for our customer the application is key why is it so because application availability performance response times also the avoidance of risk is essential for the end user experience the loyalty of the customers the revenue and so on and so on effectivity is also very important but you can optimize your system only if you know what is the performance demand of the application only then you can size the system accordingly and save maybe some money on that but who who allows all this to run this is at the end of the day the shared infrastructure and here we are talking about server infrastructure and network infrastructure we have the virtualization infrastructure cloud storage container and we have all the teams taking care of these different problems with their own tools with their own teams with their own databases and how is are these these monitoring tools are working at the end of today these these tools are working in the moment when the application is not assured anymore then an alarm pops up or in the event or pops up whatever and depending on the noise you always have you might maybe you ignore it or if you don't ignore it then immediately you start looking for it and every uh entity or every every team is looking independently for the solution of this problem even harder the whole thing is coming when a customer calls or an application unit is called is calling and tells you my application is slow please help then immediately every team is starting to investigate their monitoring tools and find out what's going wrong so at the end of the day this is an industry problem i would say we our monitoring tools are looking for performance or response times only if something goes wrong and then we are taking action and i will show you later on in the demo what is the different purpose what we are doing with turbonomic so what makes it even worse is that the application demand is changing all the time so all the applications you have are changing regarding their demand and this is this is depending on the time of day the drawing of the application so there are several reasons why these different applications are changing all the time but who's caring about the shared infrastructure so that their shared infrastructure is fulfilling their needs for the application at the end of the day nobody it takes as long as an alarm pops up and then we're looking for this whole thing what makes it even worse is when i look back in time the components are changing from my one-time physical servers to the virtual world to the cloud to the container later on so when i see that in the containerized world we have about 500 to 800 times more components we need to manage than we had in the physical or in the virtual world the next thing is the release cadence back in the days we had the annual ob annual release cadence now it's up to daily releases or even real-time releases depending on how far you are with your containerized environment the next thing is the people you are talking to in the good old days you had your iq ops but now you're getting up to the chain you're talking to the devops people you're talking also to the line of business people and everybody has something to do with performance and response time in this case and last but not least the number of management tools the number of tools every team is using is growing exponentially the more components you have the more tools you have and this multiplied by the number of teams tells you how many tools you have and to find out what's going wrong in my infrastructure what could help in here at the end of the day the whole system is based on demand and supply so the application is demanding for some resources and the infrastructure supplies these kind of resources that sounds easy but it isn't as easy because at the end of the day it's a little bit like a stacked demand and resource so for example the application is looking for more demand than the virtual machine for example or the container is giving more resources to the application but on the other hand the container or the virtual machine is asking the physical host to get more resources to fulfill the demand and the physical host is asking for the for the storage to give more resources to fulfill the needs so at the end of the day what turbonomics is doing every entity makes his decision on its own only looking on demand what's coming from the higher stack down to supply what they can supply and what they need from below so that's the basic what we are what we are doing with turbine i will show you later in the demonstration how many actions are necessary to fulfill these things and this leads later on to the question how far are we going to automate this so automation is one of the big things so far when you install turbonomics is a manual process so everything is set to manual we tell you all the different actions you can take and you decide how to take it when to take it all this kind of stuff once you come to the point that you trust the system because you have seen the actions what we what we give to you are right and improving the performance and make the whole system more stable then you have the chance to do this half manual or manual so there are different grades you can choose and at the end of the day this is looking uh this is the the supply chain what we are showing to you after all so we showing you all the different components from the apm systems for example like in stana but also other ones like dynatrace app dynamics new relic whatever you also see all the container environments you see the virtual machine you see the host you see the storage and the the example what you see over here is a very small thing this could also include cloud for example and i will show you later in the day more what more is directly in the demo how this is working what you need to know about this is no agents no configuration only data via the apis so that's all we do all this creation of this supply chain map this is what we are doing for you so this could be your infrastructure and within 60 minutes after installation we have this map on the screen and you can see the first actions so with that i would go to the demo and show you a little bit more about how these things are looking so what you see in here is our home screen i would say and immediately you see on the left hand side what we call the supply chain as already said this supply chain is a little bit bigger than the one in the example what you see in here for example we have 48 application we have the transactions the services and the application components which at the end of the day makes the application we get these information regarding uh the performance of the application from uh apm tools like instagram for example a little bit down below then you can see the containers here we have 693 containers in 462 container parts we have all the different container specs and the workloads and we have the virtual machines down below about 1000 in this example a little bit down we see the physical host we see the data stores we see the storage and this all is going down to the to the cloud at the end of the day this is a picture of your hybrid environment as it exists in your in your environment for example we can divide into the application here we can see the uh the the hybrid way but we can also see only the the on-prem view or also the cloud view we will see later in detail this is this is cool i think so when i saw this first time i thought this is cool thing but uh if this would be everything this wouldn't be the really cool thing because the really cool thing is what happens now if a customer calls you and say my application is down or it's bad has a bad response time so what you need to do and you see here my top business applications the only thing you need to know if to put the pictures a little bit higher so now you can see show me all the applications i have in my system and because we are talking already about robot shop i have a robot shop instance running as well so what we can see now we can go to robojob and immediately we see all the components which makes up my robot shop application and here you see my application transaction service components and again we have a container environment we have virtual machines running on storage so this is my robotshop application the next thing what you can see is we have these things for example by the way we see all the response time of the application because we get this information from instagna by the way you can see that it's getting from instagna because we have the small little logos over on the left hand side and this is now coming from instanta and in here we can see all the business transactions and we get all these kind of information but what makes us really unique is we have the possibility now to see the actions which we would provide to get a better performance or maybe some enhancements in the infrastructure and this is how it looks like so here we see our action center we see a lot of actions what we can do so let me skip to the resize actions for example the virtual machines so go over here so now we see that we have three suggestions or three actions and these three actions tells us we need to upsize the v memory from 512 megabyte to 1. 5 gigabyte the action category and now we're coming back again to the point performance is the key thing so to assure the performance of the application is always number one goal reduce the risk and ensure the performance i will show you later on some effectivity actions but in here so this is a performance action we can go to detail when we go to detail we see uh-huh we need to reset up the resize update memory from the virtual machine x from 512 to 1. 5 gig why because we are running at 97 over 95 of the day of the time over a 30 day observation period and this means from 97 which is way too high we are going to bring down this to 32 percent and this means we have a capacity of 1.
5 and here you can see once you would do this kind of action or take this kind of action here you can see automatically where the the usage of the women in this case would lie what you can see over here the action can be accessed accepted and executed immediately and you can do the action right from from this this screen of course you need to know a little bit what actions can i do online what doesn't hurt anything if there is something we have this concept of constraints and policies so we can give all the policies for example if this this action would need vm to shut down and restart again that we can only do this on a saturday in a three-hour change window or whatever so this is if this is done to ensure that nobody can do any damage on that of course this execute action which is now manual can be done automatically as well so this is at the end of the day this is the primary goal when i step back to my main picture here you can see now how many actions are we talking about when i show you all the actions we can see that we have about 2 700 action in this environment what we're looking for 1 800 on-prem 443 in the cloud and in the container platform again 253. all these actions could maybe be done automatically and this would keep your system always in a thing what we call desired state so we are looking for example for all the virtual machines or for the physical host that they are proceeding in a predefined desired state this could be for example a 70 percent usage because this is the sweet spot compared to usage on one side and response sum on the other side and whenever we're changing from 70 to 80 we are looking to get the the usage back to 70.