Monitoring Performance with CloudWatch Dashboards - AWS Virtual Workshop

hello everybody welcome to another virtual workshop series for our well architected framework as you know technologies evolved and certainly finding the most efficient use for our computing resources and still meeting our system requirements is important so we're going to talk today about how monitoring specifically the cloud watch but also with some other tooling is really important for monitoring and enabling performance efficiency our workshop today will be a 100 level guide so we're gonna work through some cloud watch dashboard views we're gonna look at health and you know how you can monitor your workloads and by the end of this workshop we really hope that you can have an understanding of how to use cloud watch for metrics and alarming and how to also enable into end tracing for some workloads as well so my name is Eric pullin I am the performance efficiency lead for the law architected team and so today we're going to talk about how monitoring plays a critical role in performance efficiency so as you've seen in other parts of this series we have five pillars of our well architected framework and I'm gonna focus today on performance efficiency that is also the pillar that I lead but you know what does the performance efficiency really mean and it's all about the efficient use of resources to maintain efficiency as demand changes so we want to make sure that as business and technical requirements are changed that we have technologies that will evolve to meet those needed changes so when we're developing or building or operating workloads we need to make sure that we have decisions that are made to keep the flexibility for efficiency and then we can adapt to those changes in demand and to take advantage of those new technologies when they become available so we know cloud one of the greatest features of cloud is that it's constantly evolving and that we know that there's always new features to be added and when those new features are added you want to be able to quickly consume those if they're advantageous to your workload and bring it into your environments quickly as possible so and why is this really important you know always having the best resources available for your workload is critical it will reduce your operational workload it will make sure that your systems are not only running efficiently but also maybe even more efficient than they have in the past as you onboard new technologies but most importantly it frees you to focus on your business needs take away some of the the technical aspects of your environment and really focus on the heavy lifting that you need to do for your business and meeting those workload requirements let us take on all the resource requirements as needed so it will increase the velocity to innovation and it also lead to greater business success for your company so I almost like to start by looking at some anti patterns you know what are some things that we should avoid when especially we're talking about performance efficiency one of the biggest things that I see a lot of customers do is they don't factor in all of the operational cost of making changes so for instance if you're spending months of effort to save you know one cent an hour on a particular ec2 instance that's probably not a great use of your resources you also don't want to continually impact your your workloads by you know pushing back roadmaps for minor performance efficiency games so you always have to have that trade-off of you know is the thing that I'm going after really important and will it bring the most value for my organization and you want your team spending time building things that will help your customer that's obviously the most important thing for any business but certainly when you're looking at it as a technologist we need to focus on the business perspective of things not just focus on what the coolest technology is or that we have the latest and greatest or everything we need to make sure that it makes sense for the business as a whole and so we need to make sure that we understand that the technologies that we choose do they provide great business value and the two of those things together will make sure that you are providing the best value back to your business and your customers so when we're looking at what are what are others doing in this space so when you look at performance efficiency the biggest thing that we see is companies that are continually refreshing looking at their environment and making change that environment not viewing it as a static resource but they're really constantly taking the time and effort to look through whether this is something that you know we need to make sure that we have the latest and greatest of whatever the the the technology is that you're using and so you need that continuation cycle of constantly evaluating what technologies you're using and whether those technologies are important for the services that you provide for your customers the other thing in sort of the the biggest thing that cloud provides is the ability to test new technology and so as we say removing the burden of technology what we're really saying is that you have the ability to test new features you have the ability to look at additional capabilities that you may never have had in your environment and take those capabilities and add them into your environment on a test basis before you actually commit them to production and that's really critical to make sure that as you're looking at performance efficiency in particular you you are taking the time to really understand you know what are those things that would make the most impact for you your business and your customers so you know one of the biggest things when you look at the sort of business and technology separation is what are the things that that the companies want to focus on and what are the things that technologists want to focus on and so the most critical one and I'm not gonna pick on any one technology but you know there's always the latest and greatest technology there always has been you know that's true in cloud it was true 20 years ago in IT world so you know focusing on what the latest and greatest technology is doesn't always meet the business needs and so you have to have an honest conversation with the business and with all of the business drivers you know looking at what are the things in business that you need to make sure that you're you're meeting with your objectives in your application in your workload and so that you're not taking that time away just to implement a new feature that maybe it has no direct correlation to whether your customers will be happy or not so understanding that there is this you know approach where we we as technologists always want to use the latest and greatest and I'm certainly I love technology and I love to find the latest in Grady but you have to sort of temper that with what makes the most sense for the business so moving to things such as higher level services when you can remove operational complexity those are obviously going to be great benefits to your company introducing operational complexity for the purpose of just saying that you have a piece of technology may not make the most sense so you know one of the things we always talk about is a lot of teams you know they were given technology stacks over the years and those technology stacks they had to use for everything so you know the old analogy if you give someone you know a hammer and nails they are a set of nails I should say they view everything as a hammer because they can just swing away in it but what if we can change that up and so giving your technology teams the ability to change the technology that they're using without a huge burden of what the operational aspects are that's going to be so that's one of the things that that most companies are doing this successfully they're really removing from the technology implementation teams the burden of what they need to do for implementing these newer newer technology cycles so oh it's time for a poll so one of the things that I always like to find out from customers especially I talk to customers is when you have new instance types that we release and we release them quite frequently what what sort of timeframe do you take to integrate those new instance types into your environment so we'll take a few seconds here and let everybody kind of put in the poll numbers but I'm curious what you as customers are seen today so this gives me a chance to consider water so it looks like most of you you know 6 to 12 months you're implementing most new instance types that seems pretty par for the course for a lot of customers we would like to see that time reduce for you and obviously there's a lot of mechanisms that we can put in place to help with that infrastructure as code certainly helps it when you know you can deploy new infrastructure with a simple change to what instance type that you're using you know gaining that efficiency of newer instance types and sometimes even may even change the the value proposition for a particular app workload that you're doing it is certainly critical so evaluating those new instance types as they come in but having all the tooling in place that makes it easy for you to consume those new instance types as they're announced it's critical for your business so we want to make sure that as we were announcing those things that you have the ability to quickly onboard the technology and I think that's really important for for customers as well so thanks for that poll what are some of the sort of design principles when we look at performance efficiency one of the biggest things is we want to make sure that we don't have the same text stack for everyone we want to make sure that you know every problem doesn't look like a nail you know we only have as a hammer we want to make sure that we have the ability to let teams as they're generating more clothes in the cloud that they have the ability to use advanced technologies whatever makes sense for that workload don't just leave it to when we only have a certain standard for one particular workload type you know obviously if you want up sort of the agility and efficiency in your business you're gonna have to be able to bring in advanced technologies as quickly as possible so certainly letting teams make that decision on what technologies and certainly with cloud that is much simpler because now you can utilize those technologies much more quickly the other big one is you know going global and and I know as a former IT director for many years you know the ability to implement other data centers is a sort of very time-consuming and long process so being able to take and the your workload pretty much anywhere in the world at this point with if you've done infrastructure scope with basically just a few clicks you know that's huge and be able to say that we now have the ability to go global and tell your business that and tell your business leaders that you're able to take your workloads and deploy them wherever the business may need it to be is huge and so I think performance efficiency in particular looks at how can you push this globally and make sure that you have all the infrastructure necessary to do global deployments of your infrastructure it's certainly the other half with going global is you know whatever compliance frameworks that you may have to meet so certainly you know each country may have its own unique ones and so being able to deploy anywhere in the world can help meet those business requirements as well certainly using serverless architectures is huge now we're seeing lots of customers fully embrace that and it allows you to stop focusing so much on infrastructure cost and infrastructure management costs and instead focus on what the real business value is which is performing a task for your business so using those service technologies you know just taking away that complexity of having to manage and patch and all the things that you've typically had to do when you deployed standard server workloads you certainly remove that with service architectures you also want to experiment you know this is this is where cloud is sort of been the the biggest game changer for most large organizations or really any size organization is that you have the ability to now do things and test what happens without spinning up you know large amounts of capital resources so you have the ability to really look at what is this new technology bring and what does it do for me and spend you know minimal amounts of money to do that so we want to make sure that you have the ability and you have the framework in place so that you can experiment as often as you can and really that's critical for experimentation with the agility of your work as you decide maybe this is a new line of business that we want to get into be able to spin that up and down very quickly and sort of test the ability for that is huge and so I think that you know being able to experiment vo de spearmint very often and just have that constant experimentation experimental cycle is really critical you all the restraints around what you have to do in most cases has been removed with cloud technologies so you can now operate just about any level of technology inside of AWS and we have things such as you know machine learning and transcoding and all these other sort of distributed technologies and be able to absorb those into a particular workload is key and so you can do that as you're pushing these experimentation cycles and really understand what it is that your company can utilize out of that and some of them you may and some of you may not but obviously you want to be able to take that information and then feed it back into the business so that they understand what is possible and we don't have to spend you know a whole bunch of money to do that effort so you know one of the things that you can also do is you know if you're looking at higher level services so this is another area where we spend a lot of time of making sure that people understand that they're not forced into a specific technology anymore right we have the ability to get performance changes and do things because we're not forcing that technology paradigm on them anymore you have the ability to quickly experiment look at things like let's say database platforms you know you may want to take a database platform from one technology type to another and you can experiment with that and you can see what does that really mean and so we're not forcing those decisions ahead of time because of contractual or other reasons we're going to experiment and do really what makes the most sense for our business [Music] when you look at how are we going to do this so you know we know that we need to make these changes but now how are we gonna do that we're gonna use something like a Deming cycle for this so what we're basically do is we're gonna plan do check and then act so when we first we're gonna plan what is the thing that we need to achieve and in particular you need to make sure that you understand what metrics you want to look at when you're trying to achieve newer new performance efficiencies so you need to define those metrics upfront you need to define what frequency you're gonna use those metrics and then you know how you want to monitor them and analyze those so that's the planning side of it then we're gonna actually do it and you're gonna execute a plan you're gonna test against this hopefully if you're using infrastructures code it's as simple as updating CloudFormation templates or whatever infrastructure tool you use and then seeing what changes there are so you implement those changes and then we're going to check those we're gonna study the results of that change was it a positive improvement sometimes it may not be maybe those changes didn't work the way we thought so we're gonna roll those back and if that's the case then it's very simple to roll back infrastructures code allows you to monitor different versions of that code so certainly we can roll backwards or forwards as needed and make additional changes to test that out so we analyzed those two Volt results we look for the behaviors that we want and then we look for what are our values and where those changes will and then if we say yes this is exactly what we're hoping for we reduce this particular response time or maybe we increase the velocity of a particular piece of our application then we can act on that and we can merge that in move it into production and you start utilizing that for our customers because we have planned and we know that the action will result in the pieces that we need so in some times we may say we don't we may just take that information and say no we're gonna use that next time you know we we learned something from this we learned that it wasn't enough of a performance eGain or maybe we just didn't need that extra complexity application and that's fine so we may take that and say next time we'll evolve it in a different way so you don't always have to act on those but you always want to continually look for what's the next thing that I can plan what's the next thing that I can do to reduce whatever time it is and whatever metric it is that I'm tracking is you know this is not a static thing this is the thing you're gonna do over and over and over again performance efficiency is meant to have that cycle built where you're constantly looking at ways you can iterate on that so what can you do to constantly iterate your workload to gain performance efficiency and what new technologies you can introduce that will introduce new new parameters before the workload itself so all right another poll one of the things with monitoring we see a lot of monitoring and I'm gonna talk for a minute while we're executing the poll here but you know what's the what's the area that you do most for monitoring the the CPU memory network this is the old sort of triad that we used to see in most IT organizations that everyone monitors the most monitoring tools do pretty effectively but are you monitoring for technical KPIs you know you're looking for those key performance indicators that are outside of things like CPU memory Network are you looking for business indicators so you know if you're an e-commerce website are you looking for things like the number of orders per second let's say if you are you know looking directly at how your customers are interacting so let's say you're running some kind of a SAS business and you're wanting to see what the customer experience is and how are they feeling about what you're doing so let's take a second here and see what the results are from the poll just like CPU is number one which is what we typically would see so in the low teens for everything else so 55% for CPU and it looks like 15 10% so it's great you know 10 or 15% obviously looking at those other KPIs that's critical and you you need to start defining what those KPIs are gonna be for your customers you need to have business conversations and really understand what's the most effective way that you can monitor whether your workload is effectively working for your company and your customers if you don't do that and you don't have those business all conversations this technologist I know that's kind of maybe different than what we've traditionally done but that's critically important to make sure that you have the right metrics for your organization so thanks for responding to that so let's let's jump in and let's do some hands-on lab so this is I just want to iterate to everybody this is 100 level so this is meant to be sort of an introductory view of cloud watch we're gonna look at cloud watch dashboards and a few things inside a cloud watch it's gonna give you some of the sort of basics of monitoring and you know give you some idea of how to generate and use data we're gonna be creating a dashboard and doing a few other things you can follow along with this link here and when you see this link it will take you to the wall architected labs and really this lab is just showing you how to get to we switch over to it really quickly so this is the lab this is a 100 level lab and it's really just meant to show how our Codd watch dashboards works so really simply we're just going to be logging in looking at the cloud watch dashboard one of the things I'll tell you is I went ahead ahead of time and created a couple ec2 instances you can certainly do this with if you have existing ec2 instances in your environment but if you don't then you can create a tee to micro or some other kind of instance just to test against but this lab and let me just bring this back up here again is just going to bring you straight into the cloud watch console and then we'll work from there so I've already logged into my AWS account and if you're doing this with me obviously you can log into the account that you're going to use for your testing first thing we need to do is log in and go to cloud launch you can see here from this console I have never looked at any other service so the first thing you can do is you can either start typing cloud watch here under find services or you can pull this down and then look here under management in governs sorry my screens a little big and just want my shirt so you could put cloud watch here or you can type in cloud watch which is what I typically do and it'll take you to the cloud watch console so you can see here in this console by default there we have no dashboards we have no alarms we have nothing really set up yet and in your AWS account when you initially deploy you're not gonna have any cloud watch stuff because you need to to tell us what the things are that you want to monitor so the first thing we're going to do is we're going to create a dashboard I'm just gonna call this so this is just a test that we're gonna do for a dashboard and you have the ability in our dashboards to add different charts for the metrics you have line chart stacked charts you can take number values as well but you can also do free text and markdown language so in that text field you can also have things like play books and run books so that you know how to react when certain things occur with inside of this dashboard so if we see metrics that get out of whack then we can have here's the link to you know how to deal with metrics that are in the red in this particular way you can also query log insights as well but for the purpose of this demo we're just gonna stick with a simple line chart and we're going to look at some metrics that ec2 provides we'll click that when you open up the metric graph open up a new dashboard it's gonna ask to add a metric the first metric we're gonna add here is out of the ec2 instance you'll notice here under the metrics section that we have custom metrics we also have namespaces the namespaces are meant for what are services that we're providing that you can monitor against in cloud watch so you have things like a TBS ec2 and you'll notice not every services in here these services are added to cloud watch as you consume them so if you're not a user of some of our services then you wouldn't see necessarily metrics for those until you start utilizing those services for the purposes of this demo we're gonna stick with ec2 you can look at ec2 instances in a couple different ways you can search by the ami for that particular instance type you can look at all instance types aggregated together but what I'm going to do is that per instance metric so we're just gonna look at a specific instance in this case so once I click there you're gonna see where it allows me to list all the instance types that I have and the particular metrics that they want so you can do metrics on EBS volumes reads and writes Network packets in and out as well as total bytes sieve utilization and then system status checks that you can use for ec2 auto recovery so we have these for two different instances that we have running in this account right now but I know for certain the one that I want to monitor first is CPU so we can simply type CPU in here and limit our search results to only CPU metrics the particular machine here I have two machines I have a sender and a receiver and so these two machines I'm running iperf on them as well as another tool in a second that's just gonna run a load test but basically we're just gonna generate some loads that we can see some metrics so the first one is the the sender metric so we're gonna take the machine that is the sender and we're going to create a widget for CP utilization and you can see here where the machine was spun up it was sort of set idle I ran some quick perfect tests a little while ago and it's been sitting idle ever since but we're gonna add that to our dashboard so we've added this particular dashboard widget and we can do a lot of things with this one individual widget we can edit this Widow widget we can give it a name so we know this is the sender CPU usage and we can say that's what we want to do we can change the period of time in which this metric is generated so let's just say we want to go all the way down to five seconds and see a little more granularity on that and we can update those widgets one of the other things even do is you can also dive in sort of zoom into a particular time frame so let's just say we want to see what happened during this period of time I can zoom in to that I can keep zooming in and I can see that at this particular time we reached a hundred percent so you see the utilization and then the next time interval it dropped off to 84 and then it dropped back off to the basically zero so you can zoom in and then whenever you're ready you can zoom back out again so you can see there was the initial creation of the machine obviously high CPU this machine is created and then my load tests were another another stuff set here so so we have a CPU usage for the center machine we're also going to add another widget we're gonna add it for the receiver machine so again we're gonna go into ec2 and per instance metrics and I'm going to check the search for CPU here's the receiver machine so we can create a widget for this as well and we'll change the period of time of the same this is the receiver and now we have a paired match of widgets you can see on the same timeline so the the axis the horizontal axis here is the time and so we can see what was happening with CPU so obviously machine creation on both of these high CPU we had high CPU here and we can dive into that a little bit more and you can see there was a spike in CPU but only a hundred percent CPU on the center but the receiver didn't have much of a CPU spike and that's because I wasn't running any kind of load test here I was doing a little bit of network load and just to generate some traffic so you can see here how we can zoom in we can zoom out we can see what happens during a period of time in cloud watch another thing that we're going to look at is will take network utilization as well so if we look here at EC to do another per instance and this time we're going to do the network in and out for this one we'll set this as well to set this really granular first one second and create the widget here for this is the sender machine and then we'll create another widget for the receiver machine here for instance and this time we'll scroll down and find the receiver and we'll also do the network in and out for that as well change this to one second just to have as much granularity as we can create that widget so now again we can see stack together the timeline for this of where CP utilization has changed but where network utilization doesn't change you notice here that the sender network the network out function we had a spike of almost 10 gig of data and we had a spike here of almost 13 gig of incoming network so when we're sending traffic between these two in ec2 instances obviously we're able to show that there is this traffic flowing between the two and this was something that I did earlier but we could certainly run the same instance now so if we go over here and we look at this is the sender machine that I'm going to log into and what we're gonna do is we could start a stress test and so this is just a simple Linux command to eat up some CPU time and we'll show it for about 120 seconds here we'll let it kind of stress the CPU a little bit and then as that stress happens we'll see it over here I'm going to change my time frame here to be the last hour just so we can really zoom in on this and while that's going we can also have a second terminal here and we will also do a iperf which is just a network tool that we can use to send data between two ec2 nodes just to sort of test out that we're able to see metric changes to our graphs so you'll notice here in a second as these things are starting to pop up we're starting to see network traffic is obviously increasing and we'll see CPU utilization also increasing here in a second as these graphs update so while that's happening I'm gonna let those kind of run for a second and we can show what the impact of that is in the dashboard but I also want to talk about alarming so alarms in cloud watch a lot you can create alarms based on these metrics that we talked about and remember metrics while I'm showing you built-in metrics you can generate custom metrics for any of these as well and so for those custom metrics you can do the same alarming mechanisms that we have for our built-in one but for the purposes of this lab we're gonna stick with this the the 100 level of the standard built-in metrics so for this we're going to create a simple alarm and we're going to create a CPU alarm so we're gonna select a metric here and again we're gonna go in here to ec2 and we're gonna select it per instance we're gonna look at CPU and we're going to look at the sender machine and look at CPU utilization we're going to select that metric you'll notice now under the specify metric and conditions this is in the CPU utilization metric for this particular instance and we're going to look at it over a one minute interval to try to make sure that we don't take it too much time on our workshop here and then we're gonna have a condition for this so we had the ability to use a band as a threshold so we can say if it's between a certain R and a band of evaluations that it will do that for the purposes of this we're just going to hit a static one we're gonna say that if it's anything over let's say 20% utilization that we need to fire this alarm there is some additional configuration which we're not going to do which is how many data points before you do this and then you can also set how you deal with missing data because one of the more critical things you'll notice this as you're starting to do a lot of metric collection is missing data can be just as critical as bad data as far as out of alignment data so need to make sure you understand if I don't receive data for a particular data point what should I do with that should i alarm for it or should I just go into a state of I don't have data for this and that can be very specific to the data type that you're collecting the best case we're not going to change that or just keep it the same and then we're going to take an action on this so in your action you have the ability to notify and we're going to use Amazon SNS for this so we are going to create a particular topic for this and we're just going to name this and so this will send a email to a particular email address and we can create a new one so so via test so we'll just create that I will put in my email address and we will create that topic now so what that did is it created a new topic in SNS and it's going to utilize that topic as I need to receive emails from it I'll have to subscribe to that topic in a minute and I will do that here in just a second but I want to show what other actions that we have as well so obviously with SMS simple notification service you can trigger things such as emails and that's what we're gonna do in this test but you can also trigger other things with SMS so for instance if you see an alarm state occur and you want to trigger a lambda function you can do that with SMS so you can take that s nest notification and trigger any of the SNS endpoints that are available HTTP endpoint you can send SMS messages there's you know all the features available to the Amazon SMS service you can use as a end point for a metric that has gone into an alarm state you can also send notifications when this is not an alarm state and maybe it's reached back into an okay State this is really handy when you're updating and maybe external systems that are keeping track of metrics as well so you may want to have an alarm state occur on some big board and you want to have that alarm go back to an okay state you can create a notification trigger for the alarm state and the okay state and then it will know when that's being done and it also gives sufficient data so this is we just haven't collected enough data on this as well so all those are options when you're creating notifications we've also built in the ability to take actions such as auto-scaling actions so if you're using ec2 auto-scaling groups you can do that so if you create one of these or you're using our ECS service which are elastic container service you can take that and you can deploy to an auto scaling group and say take the particular action that action could be based on the CPU load maybe we increase the number of worker nodes we have for a particular workload or you could also have it if there is a you know if the threshold drops below a certain amount maybe we say take away a certain number of nodes so you can do all that with auto scaling groups well the purpose of this lab I didn't create an auto scaling group but this is where you could set that up as well you can also set these up directly in the ec2 console for when you set up your auto scaling group so either place you can set this you can also take some actions on ec2 machines so let's just say you're taking custom metrics and that custom metric tells you something is not quite right maybe you're taking JVM statistics or something like that and you say hey I know this is a problem if I see this problem occur we should reboot the machine or we should maybe stop the instance and if it's an auto scaling group that will trigger the auto scaling group to pick up a new instance those are all options as well that you can create in your cloud watch alarm so so I'm just going to do a quick notification to an email address so we're going to do that we can give this an alarm name so this is the sender CPU to my alarm you give it a description as well and so now you'll see a preview of this so this is just showing you here's what we are creating and in fact you show the red line is the CPU utilization threshold for this alarm to trigger you can see that we triggered that alarm or would have triggered that alarm during these different spikes during creation as well as here just a minute ago and here are the actions that are occur whenever this occurs we're going to send a notification to this particular topic that we created so now we're gonna create this alarm and so the first thing you'll notice is after you create the alarm that it has insufficient data we have to pull enough to know what the data elements are for this particular arm so we're going to wait for that to occur while that's occurring we're also going to I need to quickly load up a email so I can show you here is the notification that we received for the subscription so when you generate SMS topics you have to confirm your subscription to that we don't just blindly send emails so I'm going to confirm that really quickly so I've now confirmed that subscription I could also unsubscribe if I wanted to as well and so as email alarms occur those will now come in to that particular email address so we're gonna refresh this this CPU is now at an OK state because we've pulled it enough to know that but we need to generate some more load so that we can have this so let's put another stress on the CPU now one of the things to note about CloudWatch alarms is we we have to see this particular condition occur for more than one data point and because our data points the granularity is within a minute we're gonna have to wait between a minute and two minutes it just depends on exactly when this gets triggered for this alarm to occur because we need to verify that the condition has been met for that particular time for it so we'll clear that out and we'll let that go while that's waiting to be triggered and we'll see this alarm come up here in a second one of the things you can do is you can also add a widget for the alarm I mean I'm going to take this alarm and I'm gonna add it to my dashboard so you can see here I have the only dashboard I have created I'm gonna create this in there what i named it the same thing that it was and i'm going to generate that and now if i go in here to the dashboard not only do I see the CPU utilization now I can also see this is the threshold for the CPU zation alarm to occur if I click on those I can dive in a little deeper and see what's going on you can see obviously in the last hour we had a spike in traffic or a spike in CPU utilization that was due to my stress test we're seeing another spike here occur and right now this is not an alarm state but after we do a second pull cycle you will see that that will suddenly jump into an alarm state and when it does it should trigger an alarm for us to see in our email as well while that's occurring you can still as before we can rearrange these so if you want to take widgets and let's say hey we already have a one of these widgets we can remove that so let's just say we want to delete that one and now the widget that we have that is the alarm widget which also includes CPU as there now you notice is refresh this we've now exceeded our threshold we have an alarm state that's occurred on the dashboard itself we now see a red box around the alarm and a little exclamation point there and if we go over here to our email and excuse email sometimes it can take a minute so we will wait for that to refresh and we should see an email come in as well you can set alarms on any of these so if we wanted to set alarms based on network traffic we could set it on CPU and also any custom metrics that we do we also can collect metrics as now with our caught watch agents so you can collect metrics on any number of items that you want to configure on the machine so things like mem feel memory which is obviously critical for some of our workloads and those things or even the custom metrics of things like your e-commerce orders per second that would be a great metric to have as well so and here's the alarm so you can see that I have now received a message zoom in to make this rule easier to see this so you have now received an alarm and it's showing that the sender CPU is too high for our instance across the threshold gives you some information about the alarm detail then obviously we can also click into this alarm directly within the console and as the CPU winds down and our run is completed so it's run for two minutes it will take another minute or so for that alarm to do away and you can already see Sabula zation has already started to dip down so if we kind of zoom in here you can see the last data point that occurred it was at 68 it'll take another cycle or two for it to fully go down to 0% again and then we'll see that this is no longer you know State so that is the the general look at CPU metrics and and other metrics that are generated for you obviously most of our services generate metrics things such as like RDS you can get information is the number of database connections or you can get information about things such as you know CPU ization memory utilization of the database engines so you know most of our services do provide those metrics out-of-the-box and then the custom ones you can you can generate as needed so wait for this to fall back off and as you can see here it has fallen off again so we're back down to 0% CPU we look at the alarm you'll see that it's still an alarm state and the reason for that is we haven't had two consecutive data points we've only had one so when the other data point occurs we will see this go and has now gone green so if we refresh this is green and is now in an OK state so we have now made sure that our alarm is back into an OK state so with that let me flip back to our presentation that is the lab that we have for COD watch dashboards but we're also going to look at something else related here to in a minute alright so that was our so one of the things that you just to give me an example of what you can do with these carwash dashboards this is an example one this is looking at a web app you can look at things like load balancers in this case so the response time for an alb you can look at you know see the utilization and with this dashboard you can see that you have the ability to put all those pieces together to see the totality of your workload you know are we concerned with response time or are you concerned with network traffic are we concerned with database latency you can look at all those in one place for that workload and have a dashboard that lets you view it over a period of time and so as you can see here this is an example where we saw CPU utilization Speight which also means database latency it spiked which also means our auto scaling had spiked as well and the number of connections also arose so all these things work together and you know that looking at these different mechanisms for metric viewing that you can say we see what the problem really is maybe was caused by the CPU utilization going up everything else was just a downstream event from that so something else that you have to think about when it comes to performance efficiency is observability so how can we help you observe what's going on with your workload and get better insights into what's going on you know for many years most of us have built and run monolithic applications and that was really for a lot of companies even today that's the start of your cloud journey so you're making that move from the monolithic to more agile micro services based architecture and so you may be at the beginning of this journey you may be in the middle or even at the end of this journey but anywhere along the way you are having to have the information regarding the availability and the performance of that application as a whole so when we looked at you know the monolithic applications you know how we see it was a lot for one team to manage and this is why you know the complexity of making changes it would take months in some cases years to make changes to production environments because there were so many changes so many teams involved with changes so you you would have to unfortunately that would slow down your ability to innovate and it was slow down your ability to sort of reduce you know any of your customer concerns or maybe introduce changes that would help differentiate your product from others it's so model of the applications unfortunately just don't allow for the sort of the quick iteration that you're looking for so as companies have transitioned there are working into breaking up their monoliths into separate services and those services may be owned by different teams this allows you to be much quicker so your agility of manipulating your workload increases the velocity of that and this is obviously the same journey that we've been on here in Amazon we've been moving from monolithic to micro services for many years and and you know it allows us the agility to make changes very quickly but when you move to this it obviously introduces a lot of flexibility it also allows you to free up resources to use things other than the one stock standard application portfolio that you may have had before so for instance if we look at a particular component that we need to build for our workload and that component would be better suited to be built in a different language by decoupling all of these into separate micro services we have that ability we can build something maybe the base application is built in Java maybe we need to build something and go or Russ and that's what makes sense for that you know killer new feature that would really grab your customers attention by building these as independent pieces we can certainly do that and you can you don't have to change that entire code base to make that happen you also have when you decouple the ability to change the way that you view the operations of this so when you're looking at you know the way you're gonna operate and you're gonna run you may have some sections of your code base can obviously stay on ec2 instances in this cases VMs you may move some to higher level services such as RDS we take on the operational efficiency part of that you may also look at things such as lambda where you don't actually need to have the operational complexity of running the workload on a machine so you can certainly take in the couple of years services and you can also look at multiple different ways to run those services as well but when you do this the one critical piece of this is we start decoupling and you start having this distribution of stuff you need to have observability across all that for the entirety of your application so when we look at you know you lose visibility I should say you lose visibility you struggle with visibility when you have these large decentralized systems so you have to bring in tooling that allows you to see them across all the various components that you have so let's just say as an example here your application your workload as a whole you may have some degraded state where you have a spike in latency and your application goes into that degraded State you know you need to be able to quickly pinpoint that and how are you going to pinpoint that and you need to be able to find where along the way is that latency causing you the pain and so being able to put all these pieces together it's gonna be critically important as you do this decoupling exercise so one of the ways you can do that is by looking at Indian tracing so AWS x-ray service allows you to take your application and generate a service graph and so with that service graph you can take each a devere's resource and you can look at how they interact with each other so it takes these really complex services and sort of simplifies it to the point where you can look at each individual component look at what their reaction is from each one where is latency a problem maybe and then see that in a totality of the service graph and the AWS console is an x-ray console will show you and even generate this visualization for you it also introduces the concept of traces so if you want to be able to take a particular session and a particular thing that a person or another system has done you can generate a trace ID number and then it will track that throughout the entirety of your system so it will collect things like the HT begin or post it'll show all that traveling through the load balancer it'll show it hitting your application code it'll show any downstream services you may be calling so you know things such as external web api's or maybe you're hitting even internal API is that you've generated and it'll show you the time frame for each one of those inside of the X ray console and it also includes the idea of segmentation with inside of that so you can look at your computer sources running as each bit of work they do you can view that as a segment and this segment shows the resource name the details about it what work was being done for example when you have an HTTP request reaches your application you can record where host it comes from the request the response what work was adult was done and then also any issues that may have occurred it was a four hundred or something else where you you get exception stacks for that so x-ray allows you to view those compute resources as different segments along them along the path and really best of all X ray tracing is pretty simple to enable especially for lambda and API gateway for lambda it is just a simple you know checkbox to enable active tracing if it's API gateway it's the same there and so now you can suddenly view the application the the the different calls through the application through an API gateway obviously we have workloads that don't use those two services and so you can also enable X ray tracing directly within your code so when you do that you can just in this case we're gonna take the AWS SDK and we're going to also require the AWS X ray SDK so we can see all those AWS API calls layered on top of that so so calls to things such as DynamoDB or even other higher level services and you can do that we support this for a variety of languages and you can use look at the x-ray home page to see what languages we support today and so we have a load of metrics and cloud watch we have all these things in x-ray so let's let's take a quick view of of what x-ray is and just kind of give you an idea of what this is this link is to a lastik beanstalk sample app but i'm also going to show you in the console how you can quickly do this with x-ray so if we go to x-ray in the services and the console you can see this is the x-ray console here and we're gonna get started oh I'm sorry let me in the show here's all right okay here we are let me cancel let me go back down here so if you are in the management console and you can search for x-ray you open up the x-ray console you may see that getting started at the screen if not you can click up here and getting started what we're going to do is we render deploy a sample application the sample application is going to show us how interacting with the different components inside of AWS happens with x-ray so this sample application is an elastic Beanstalk app it uses nodejs on the front and it's going to build this so I'm going to go ahead and launch this and if you're doing it along with me you should be able to do this as well this is just again a confirmation template we're going to launch it into our V PC this will take about two minutes for this to complete so if you just go so I'll let that run the the x-ray sample is going to spit up an elastic Beanstalk cluster I see Beanstalk cluster will hold the application and we'll get to see x-ray implemented so that it it also uses a database table and dynamodb will see that it also reaches out to SMS topics and so you'll get to see the the totality of utilizing x-ray for that so while that's being done we will wait for the service map to start up again we don't have a service map yet because the application is still spinning up so one of the things I'll show while this is spinning up is whenever you're looking at the x-ray application and we'll see what conformation is doing here yep so it's spun up an elastic Beanstalk and we'll see that spin up here just a second it also created some IM rolls and some IM policies [Music] as with with conformation we'll just have to wait for it to spin up those resources in the console once we get our initial configuration we'll be able to see the service map of the way the application is we can look at those traces and we also have an analytics component now as well [Music] so let me switch back while we're waiting for that to spin up I think there might be a good chance to do our next poll and we'll wait for that as we're waiting for confirmation of finishing up so are you currently using a monitoring system for Indian requests or is there some kind of tracing system you're using do you know about that and and is this something that's widely used in your and your business today this is something that I think is critically important I know for many years we have you know spent a lot of time on our CPU you know the sort of the Trinity of CPU Network and memory utilization and we have a lot of monitoring systems for that but I think a lot of people are you know starting to see the value in having into end monitoring so you're not just looking at those you know sort of ancillary metric data points and so the what the customers actually seeing so looks like the majority of you are saying no so it looks like but 66% are saying no 17% yes 16% unsure so this is a piece that I think you know as you're growing into your cloud environment that it's it would be really advantageous to look at how you can do end and monitoring obviously x-ray can do it there is other tooling as well but you know it's something where having that ability to see what all the components along the way are doing will be critical as you're decoupling reducing those model lists to individual components it's it's a critical piece of that so alright let's flip back and see we're almost done with that so what you'll see here is confirmation is spun up the initial x-ray sample and then elastic Beanstalk is doing its work what elastic beanstalk is doing is it's actually creating the various component pieces that it needs to run this coating auto-scaling groups so the actual machine to run it so all that's still going to be working as well let's see one of the things when this application is complete we will see a output section here we will actually have the IP address of the machine that we're going to hit to see the front-end console to run this test so we're gonna have to wait for that anytime you're doing this stuff live it takes longer than you expect so we'll have to bear with it as it's creating its auto scaling group x-ray is you know one of the one of the services that we utilize with a lot of customers that just don't have visibility today it's is relatively simple to implement and I think it's one of those pieces where you know customers can gain a lot of value out of their sort of the simplistic view of what the application is doing and kind of get a better view of what pieces and parts that it's talking to so you know if you haven't evaluated x-ray I would I would highly encourage you to do that looks like we're almost complete here so we're just waiting for the outputs to be to be triggered if you're doing this along with me it may take a couple minutes as well for to spin up and and the end of this we'll want to make sure that we spend these back down and so we can use confirmation to do that we will take this stack right here we can just delete it and that will that will delete our stack for the x-ray sample but this simple nodejs app is really just a way for you to view what's going on okay now we're complete so here is the when you go to the output section for this x-ray sample you'll see this elastic beanstalk environment URL this is an IP address and obviously it's already created all the components that it needs to run this application this was also done in the default V PC and that was meant so that it would have internet connectivity so if you're deploying this in a internal V PC you'll need to make sure that it lives in an environment that has the ability to get out through an Internet gateway so with that said if we bring up the actual website for this you'll see this is the x-ray sample application and this application has just really sort of meant to be a quick and an easy way to see x-ray in the console so I'm going to go ahead and start this up why I explained it so what this is doing is it's a sign-up page so it's a really simplistic signup page that you can have if we click sign up today we'll ask for your name your email address and then if you're interested in this preview access and you say yes and it says ok that's all it does on the back end we're actually storing this data and talking about dB we're using SS topics as well well you'll see is when I click this start button at the top I'm actually just generating load so all this is doing is hey I'm just signing people up so you can see here I have 6 requests that have come through so far and this is a 7 so I'm going to stop that so we've had basically eight requests come through the system and now we can go into x-ray refresh the x-ray console we can see the service map so the service map is being computed for this particular application and now we can see what this application is actually doing so we have our end clients here we're coming into this ec2 instance you can see that it's hitting three remote services the first one here this SNS service that's where we're sending things onto a topic we could actually email individuals through the service if we so chose we see the average time it took to do that we also have a dynamodb table and you can see that this dynamodb table is being written to and then this one here is interesting because this is our local instance metadata so one of the things you'll notice is we whatever the service is spun up for the first time it also has to reach and get credentials to access things such as dynamo d USNS so on the ec2 instance it has to access that local instance ability that is also shown here in x-ray as well so we can look at not only as a map view and can I see what the overall health is of each of these components but we can also do a trace setup so we can look at some different traces of you know here that you uh URLs that were hit we were okay on most of these if I scroll down and look at individual ones I can see that you know here is what a get request was where it just got a particular instance I can dive into this particular segment and see when it was started what happened you know what was the actual request URL where they come from all that information where's the ec2 instance that actually managed this request all that information is available in the trace details we'd also look at a post method and this will give you a better idea of the actual indicator of saying that this person wanted to sign up so you can see that we have this front-end application but this front-end application also needed to hit DynamoDB and ordered a hit DynamoDB and needed to ask for credentials and so these credentials it got from the local instance metadata in fact we can even dive into that we see here that this local instance metadata you can see that what it requested was the security credentials for accessing this particular service so we're able to see that if for some reason you know getting the local instance metadata was taking too long or how long did that take we can we can dive into that level of detail with this we can also see you know how long did it take to actually write that particular bit of data out to dinamo and then what was the actual item that was put all that information is available in the trace as well so that's just a sort of a quick overview of how to use x-ray and how the service map can be used and again this is a sample application but if you enable this for your application it will try to define all the various component pieces that are utilized and show you traces for all those in order to spin down this lab you can just go back into cloud formation and click on the x-ray sample scroll back up to the top say delete we will delete this stack it'll take a few minutes for it to delete as well but it will delete hit and the elastic Beanstalk app that goes with it that way you're not being charged for this particular application past the demo so with that said let's get back to our presentation here and so when we talk about performance efficiency and we talk about what are the questions that we need to ask this is you know obviously this is part of our wall architected series and I'm the performance efficiency pillar lead for that and so there's a list of questions that we ask as part of well architected that leads us to whether you're being wall protective as it relates to performance efficiency so let's just look through some of those and I'll kind of guide you through them if you use the wall architected tool which I'll show here at the very end but these questions are embedded in the tool and so this is what the tool helps you with walking through these different questions and understanding them so first question is how do you select the best performing architecture we we know that it takes multiple approaches to get optimal performance so the the best way to do that is to understand all the various component pieces that are available you have to know what services and resources are available to you to utilize so that you're utilizing the correct ones for your particular workload you need to really have a well-defined process for you know selecting those architectural choices is there constraints that your businesses put around due to regulatory concerns or other things so obviously factoring in cost and then you know maybe even utilizing external partners to help guide you on what the best way to implement or what best architectures for your particular type of workload and then I think the most critical thing is to benchmark what you're currently doing and so you know what you are expecting today so that as you make changes in the future you know you have a baseline to work against when you're doing additional load so obviously the first thing is how do you select that best performing architecture you need to select the right compute platform you need to know what compute platforms are available in the cloud you need to select what the most with the best performance compute workload you need for your particular workload and so traditionally you would see a lot of customers tend to just err on the side of really big and I will say that that's well that's perfectly fine you know there's certainly a cost component and we also have our cost pillar that goes into some of those but you know in order to make that decision effectively and be at the the right cost you need to also make sure that you're collecting any of those compute related metrics so that you know that the the selection you made make sense there's nothing worse than having a machine that you're paying you know a fair amount for per hour that you're using 1% of its CPU it just doesn't make sense so you know select the right resources making sure that you have the right metrics related to those resources so that you know the thing that you're selecting is the correct thing and then obviously right sizing is part of that where you just would say hey these are the things that I can do to make this smaller or bigger as the workload did taste and then huge is to change workloads to be as elastic as elastic as possible so you need to make sure that your resources have the ability to grow and shrink as needed based on your workload so constantly re-evaluate those make sure that those metrics make sense for you and then use those metrics to make decisions on what compute resources you're going to use same thing with storage make sure you understand what your storage requirements are evaluate what configuration options you have available and then make sure that you make decisions based on that so we have a lot of different storage types we have everything from object storage and s3 to a FS to our FS x products those are all different products that you can do with a block file object you name it we have file systems that can meet most of those demands and you need to make sure that you may use even multiple stories and each component may have different story services needed to be the most efficient and it can be a database solution this is where we see a lot of customer spending a lot of time especially in their initial journey day to the cloud is understanding what performance related characteristics and other data characteristics around their database solution and then looking at what options are available RDS is a great option for a lot of customers where they can get rid of that operational sort of heavy lifting that they've had to do in the past and refocus themselves on the things that are important such as what's the best way to index and are we doing the right tablespace for a particular database engine type so looking at what options are available other than just running it yourself is is also big and then again collecting those metrics you also need to think about data storage when it comes to database solutions and then also looking at other database solutions as that are outside of the standard ODBC style systems you know is this particular component maybe a better fit for a no sequel database such as dynamo or is ElastiCache with Redis a better option those are all things that you need to evaluate you know in traditional IT we typically stuck with one database and the reason for that was we didn't want the operational complexity but with cloud technologies we can remove that operational complexity of those other services so why not take advantage of those if it's the right thing for that workload you know how do you configure a networking solution so this is another critical piece what are you going to be doing with how network impacts performance what options are available for connectivity so you need to make sure are you connecting and on-premise database to cloud are you connecting you know other data providers you know what what are your constraints related to understanding what options are available from a network perspective you know understanding what options are available VPC options that we have and and what is necessary for that there are also some best practices around things like you know minimizing the number of Knakal rules and and how to utilize I'm sorry Knakal Network address our network ACLs I'm sorry but also how to compare and contrast those with AWS security groups because that is a slightly different concept that maybe what you traditionally have done also selecting the right protocols you know looking at we are talking about high efficiency versus you know high reliability is that something where you need to choose different protocols based on the workload type and then choosing locations obviously we have regions throughout the world and we also have front-end Cod front distribution pops that we can use as well so those are all things where you can choose the proper location for making it best for your customers in your workload and then constantly evaluating the metrics related to your network we provide lots of metrics inside of AWS for you to consume and then you make the right choices for what the networking solution needs to be and then as you collect all these metrics you understand those you need to have a mechanism internally to work to evolve your workloads over time you need to have the ability to keep up to date with what new resources and services are available you know we're releasing new services all the time and it's important for you to to understand what those services are and how they can impact your workloads you need to define a process on how to keep up to date with those and then you need to define a process for constantly evaluating those workloads and what the performance characteristics should be as you evolve your business your workload may also evolve and so the characteristics of what you're worried about for that workload may also have evolved to constantly reiterate what those workload characteristics should be and then obviously just evolving the workload over time from a performance characteristic standpoint may be you know when you had just a few customers that the performance characteristics were much different than now that you have an established product and you have you know hundreds of millions of customers so those are things that you need to evolve over time as your product grows and changes and then how are you gonna monitor all of this you want to make sure that you know system performance it can degrade over time so you need to make sure that you have the ability to monitor that have the the understanding of when something gets degraded you know what is your factors for that degradation and then how you're gonna remediate that so you know is it a operational lobe problem is it a configuration problem those are all things that you need to have all those metrics in place to understand what the different options are so you need to analyze those metrics what events occur and then you need to establish KPIs to constantly measure the workload performance what are those key performance indicators that you need to make sure you know reviewing metrics at regular intervals and then obviously as we show today monitor and alarm on those proactively and then trade offs you know one of the most critical things that you have to do when you're valuing technologies is where the trade-offs at so when we're looking at trade-offs for you know there may be it increased complexity but we may get a you know faster performance or is there a pattern or change to our service that we need to make we need to evaluate how those trade-offs can occur and that there's also a cost trade-off you know we can have the most performant system in the world but it may cost a lot of money does that make sense for your workload you need to measure and understand if that impact makes sense or not you know if it's too expensive to run and your customers won't consume it then maybe it's not the right performance that you need for that so and then obviously just using other performance related strategies you know trading things like are you gonna trade consistency for durability or space for time and latency those are things that you can trade off that may impact the way you design an application and a workload but otherwise it you know it may make the most sense for that particular work so those are the eight questions in the performance efficiency pillar so now we'll ask our last poll question so now you've kind of heard some of the performance efficiency questions you know what are what are questions that you think would be the most interesting for a deeper dive on obviously there's the selection process looking at what metrics to monitor and what technologies to use reviewing those metrics or you know monitoring of those metrics or doing a deeper dive on how to select trade-offs for your workload so we'll let that poll go for a few seconds so it looks like a pretty overwhelming majority said well 44% think that monitoring is a critical piece 24% it wanting to know more about trade-offs and then selection review not so much so 18 and 9 percent for those two so wait for one more refresh here and see what that comes back yeah so the vast majority of you want you know deeper dive into the actual monetary task and I I think that's a fair ask and I think that most companies that's where they're struggling now and so technology such as cloud watch and x-ray are great examples of things to dive deeper in and as you look at a future you know workshops those are things that you can focus in on to you know sort of add to your toolbox of things that you can implement for a particular workload at your at your company so it's great so with that that's the end of the presentation for today and we will go to Q&A now okay great so let's let's see what questions we have today so question how does x-ray collect Reese's yeah x-ray clucks traces a couple different ways you know you can inject the SDK directly within your application and depending on the service if it's lamda or API gateway it just injects itself directly into those API calls to give you information on that [Music] the questions we have here don't have any questions other than that one so far does anybody else have a question that they would like to ask do you recommend ETS or Fargate as it relates to which is easier to monitor retrieve logs from I would say that I don't have a recommendation for one over the other as far as monitoring and retrieving logs I think you can do on either of those so I don't think that I have a real preference on one or the other I think the thing you need to ask yourself is what what level of infrastructure do we want to manage you know when it comes to eks versus Fargate I mean you know kubernetes versus you know what we're doing with our gate and ECS is there somewhat different so you know I think that's a fundamental question needed answer but as far as monitoring those we have monetary built into both of those and we can you know certainly es you can do Griffin ax and take those logs and dump them into directly in a cloud watches custom metrics there's a lot of options and then ECS we have some of those directly built into cloud watch how to monitor business KPI so this is an interesting question there is obviously you you have to have them the mechanisms in place in your workload to monitor those so you know the example always uses orders per second so that may be a metric that you're gathering through some kind of queuing system or maybe you're gathering it through a database table but deciding how to gather business KPIs is is a critical piece to be quite honest and so you need to understand what the first you have to understand what business kpi's you want to measure and then working backwards from that to say now that I know these are the business KPI is I need to measure how's the most effective way for me to gather those KPIs and I think that's really the the critical piece is understanding which kpi's you want first and then understanding how to work those into the application so you may absolutely have to generate those Bethan business metrics you may not have those today so that may require code changes and other things so every webinar ends with monitoring and it's resolution how about analytics because I don't quite understand the question [Music] [Music] how do we monitor application logs okay we didn't go into that today but you know with AWS cloud watch logs you can you can take application logs directly into well directly from your application and it into AWS cloud watch logs and then you can monitor those and actually report against those and then have alerting based on those you can also trigger events based on log metric filters so again we didn't go into that today but those are options that are available as well [Music] looking for some of the ones Oh so how do vendors compliment or compete with cloud watch well I would say that you know Cobb watch is is certainly a system that we use internally but most of these APM vendors that we have they directly integrate with cloud watch they consume logs from cloud watch and then present it along with other information so a lot of these vendors will you know take real time user monitoring and and other things that put them together and in their own unique ways and those are great you know cloud watches is built in AWS but there are certainly others that can that can complement those so there's definitely no competition there it is certainly a cooperation between those [Music] yeah so the Analects question was about using the data to perform analytics later so yes the now called watch metrics in general are always stored for a I think it was two weeks period of time but you can take that information and you can put it into external systems to utilize over time so you can take it and put it into you know a redshift cluster if you want to analytics against it if you want to dump it into some other kind of data warehousing system that you have you know those performance metrics and those things you can certainly take those as well so ah so can you talk about providing visibility to upper management for monitoring is there solutions for that so certainly with cloud watch yeah that may be a little bit too deep for some you know sort of sea level if you're trying to elevate those up but don't lose sight of the power of some of those metrics and the graphs that go with those metrics so yes those may be a little too deep but you can still utilize the data out of those metrics and they actually charts themselves in other presentations that you may be giving so showing correlation over time that you know a particular event occurred and that event caused these downstream services can be really powerful to show all those things together in one sort of view you know that you know it wasn't just that we ran out of computer or that we had a scale-up compute it's that we ended up you know we caused these three other downstream services to cause issues so you know those are all things that that certainly I would utilize when you're talking to upper-level upper management last question any plans on any blame quick site on cloud watch logs I don't know of any I'm sorry I'm not the product owner for excite or cloud watch but you know that's something that certainly you can reach out to your local AWS representative and see if they have any information about that I think that was the last question so with that I think that we're done

Channel: AWS Online Tech Talks

Views: 26,314

Keywords: Performance Efficiency, Well-Architected, AWS, Webinar, Cloud Computing, Amazon Web Services

Id: _eWaPVski6o

Length: 86min 16sec (5176 seconds)

Published: Tue Nov 12 2019