Automated distributed services tracing is key in microservices environments to uncover performance and functional issues. Dynatrace provides automated tracing as cloud native platform feature.
WSO2Con2024 - Software Delivery in Hybrid Environments
Automated distributed tracing - a first class citizen of monitoring
1. Confidential, Dynatrace, LLC
Automated distributed tracing – a first class citizen of monitoring
Alois Mayr, Technology Lead at Dynatrace Innovation Lab, @mayralois
May 23rd, 2017
13. Confidential, Dynatrace, LLC
What does health mean for App teams vs. Cloud Ops
Microservices Team A
Microservices Team B
Microservices Team X
Dev Test/QA Prod
Cloud Ops Team
Application service insights, performance, failures,
scalability, interoperability
Platform health, capacity planning,
Monitoring as a Service
Dependencies,
impacts,
architecture,
end-user-experience
14. Confidential, Dynatrace, LLC
Monitoring as platform feature
host
Deploy one agent per host (kubectl create, bosh deploy)
Auto instrumentation of containerized microservices
Auto distributed transaction tracing
AI based root cause analytics
Cloud Ops teams own the platform
15. Confidential, Dynatrace, LLC
Container health != service health != application health
Container health
Microservices health
Application health
IaaS health
16. Confidential, Dynatrace, LLC
Container health != service health != application health
Container health
Microservices health
Application health
IaaS health
Chaos Monkey
17. Confidential, Dynatrace, LLC
Container health != service health != application health
Container health
Microservices health
Application health
IaaS health
docker stats
18. Confidential, Dynatrace, LLC
Container health != service health != application health
Container health
Microservices health
Application health
IaaS health
19. Confidential, Dynatrace, LLC
Container health != service health != application health
Container health
Microservices health
Application health
IaaS health
monitroing used to be about looking at dashboard. You developed you application, deployed itin our enviornment either on-premise in your DC or in the cloud on AWS for example.
And than you watched if anything happened or if any set threshold throw an error message. But modern monitoring should go beyond that…
and to make this a bit more intuitive, lets launch Dynatrace.
In that Demo environment we are simulating to have multiple applications running, incl. customer facing ones. Have of the applications run in the DC, have of them in AWS.
So what Dynatace does, it provides you an full stack overview of your environment – starting from the underlying infrastructure up to the real user monitoring.
We show you all dependencies within your stack, with SmartScape we basically auto-discovery your whole environment.
Now assuming you just start your cloud migration journey, having SmartScape in place makes the whole profiling and discovery phase very ease.
But simple installing our agent in the environment your AS-IS landscape will be created –out of the box – within less than 5 minutes.
Now you can use this information to start the actual migration process – whatever strategy you are following (lift and shift, refactoring, replatforming) and as soon as you have migrated the workload you need anyway a cloud native monitoring solution.
Product demo – lets check this out in the real product:Assume we have following scenario. Have of the applications are already running on AWS, half ot them in DC.SmartScape – full stack, auto dependencies, auto discovery,Cloud Migration Use Case – how to support those
The typical life of a container is minutes and hours and not necessarily days. Given this kind of environment, it becomes difficult to monitor the applications as the number of nodes are ever changing.
As enterprises migrate applications from monolithic to microservices and use containers to support them, the number of containers supporting these microservices explode over time to 100,000s of node. A typical monitoring is unable to handle the amount of data generated by these 1000s of nodes.
monitroing used to be about looking at dashboard. You developed you application, deployed itin our enviornment either on-premise in your DC or in the cloud on AWS for example.
And than you watched if anything happened or if any set threshold throw an error message. But modern monitoring should go beyond that…
The typical life of a container is minutes and hours and not necessarily days. Given this kind of environment, it becomes difficult to monitor the applications as the number of nodes are ever changing.
As enterprises migrate applications from monolithic to microservices and use containers to support them, the number of containers supporting these microservices explode over time to 100,000s of node. A typical monitoring is unable to handle the amount of data generated by these 1000s of nodes.
The typical life of a container is minutes and hours and not necessarily days. Given this kind of environment, it becomes difficult to monitor the applications as the number of nodes are ever changing.
As enterprises migrate applications from monolithic to microservices and use containers to support them, the number of containers supporting these microservices explode over time to 100,000s of node. A typical monitoring is unable to handle the amount of data generated by these 1000s of nodes.
The typical life of a container is minutes and hours and not necessarily days. Given this kind of environment, it becomes difficult to monitor the applications as the number of nodes are ever changing.
As enterprises migrate applications from monolithic to microservices and use containers to support them, the number of containers supporting these microservices explode over time to 100,000s of node. A typical monitoring is unable to handle the amount of data generated by these 1000s of nodes.
The typical life of a container is minutes and hours and not necessarily days. Given this kind of environment, it becomes difficult to monitor the applications as the number of nodes are ever changing.
As enterprises migrate applications from monolithic to microservices and use containers to support them, the number of containers supporting these microservices explode over time to 100,000s of node. A typical monitoring is unable to handle the amount of data generated by these 1000s of nodes.
The typical life of a container is minutes and hours and not necessarily days. Given this kind of environment, it becomes difficult to monitor the applications as the number of nodes are ever changing.
As enterprises migrate applications from monolithic to microservices and use containers to support them, the number of containers supporting these microservices explode over time to 100,000s of node. A typical monitoring is unable to handle the amount of data generated by these 1000s of nodes.