It’s the new new! The DevOps universe has come to agree, the new world is about constant change. At the infrastructure layer, containers running on virtual hosts that come and go based on processing demand. At the application layer, micro services that align precisely to the containerized approach create a world where the infrastructure is servicing the needs of the application and the architecture is morphing constantly.
This is all great for the developers! They can now build up new business capability without worrying about silly things like capacity. But all this dynamism creates new challenges - operational quality of service. Unfortunately, just because we adopt these new techniques does not mean that it all works all of the time!
Monitoring and logging have always been the feedback loop for DevOps to understand quality of service. But traditional monitoring tools were never designed for monitoring components that by definition disappear, much less have a life cycle in seconds rather than days. Scale is also an issue, containers create a world that naturally have 1000s of components to understand.
Here at Instana, we are creating a new approach to managing modern applications. We have created Stan to be your devops assistant. Stan is here to help you understand where problems are, and what you need to do to fix them. However, Stan needs a pallet upon which he can help you visualize your situation (components, component health, dependencies, application architecture, and finally service health). With so many moving parts, and no master plan on how they fit together, visualization of architecture is a critical need. The requirements for visualizing modern environments include:
- immediate visualization - new components must appear in the system hierarchy in seconds
- dependency - any component has suppliers and consumers, where does it fit in the graph?
- impact - when a component’s health is compromised, what is the downstream impact?
- scale - 1000s of components are quite common, what is the strategy to visualize this scale?
We have run a large scale simulation and created a video to demonstrate the Instana approach to these requirements. Please take note of a few details:
- real-time rendering - when a server is instantiated, details are discovered and visualized in seconds, likewise when a server is de-provisioned, it immediately is gone
- signal through the noise - when there are 100 - 1000s of servers, how do you spot the offenders? Yellow and red on a sea of normal grey servers is quite clear
- grouping - in modern architectures, components are rarely alone and usually clustered, this is easily visualized on the dynamic graph
- tagging allows the customer to name their own grouping models, but simple examples are shown
- Dependency - Instana dynamically maps the interconnections and dependencies between components which enables impact analysis and real root cause determination
- specific dashboards give contextual information designed specifically for the unit under management, not just reams of data
Please visit our Youtube channel and watch our latest video: