The three pillars of observability have proven themselves a crucial way to trend, tune, and troubleshoot systems large and small. These are critical day-to-day operations in millions of organizations, as attested by years of collective infrastructure management experience. But as globally-distributed infrastructures become more common, are they enough?
The Three Pillars of Observability
Just to make sure we are all on the same page, the three pillars are: logging, metrics, and tracing. Each pillar provides a solid piece of the puzzle
- from seeing warnings and security alerts in the logs that infrastructure components are producing,
- to tracking metrics around network switch capacity and disk I/O,
- to stack traces in Java applications that show the exact place in an application where an error came from.
Together, the three pillars provide a real feel for what is happening inside every service that is running in your infrastructure.
Yet we no longer live in a world where services are stand-alone. We can’t gather metrics from every step on the path – from the client using an application, to the disk array in the storage area network the database is writing to.
Application Performance Management (APM) products have been on the forefront of building on these core capabilities. Leveraging the compute power on each client device brings additional visibility to the complex and distributed environments that are becoming commonplace.
User Experience Monitoring are Metrics++
“Distributed systems” does not just mean that the services and servers are spread out over multiple clouds. The individual users that consume the services you offer are increasing in both their number, device of choice, and location. This can range from remote workers connecting via VPN for their full eight-hour workday, to a client using an app on their Samsung Galaxy to pull up the tracking number from their most recent order while waiting in line at a local UPS store.
Having visibility into the client’s actual experience while using the application – i.e., observability – is now a vital part of the metrics required to operate the applications in your infrastructure. The most-used methods to gather this extra data are introducing a JavaScript library into web applications or a custom module into a mobile app. The metrics gathered provide everything from network latency, to screen sizes, to operating system and browser versions, and even what plugins and languages are available.
These metrics alone provide valuable insight. They help to find issues such as a specific region having slowdowns. They help to identify where adjustments can be made to infrastructure components like adding more content to a content delivery network. The three observability pillars make it easier to find potential problems before clients notice the impact and start calling in with actual problems.\
Topology Allows for Tracing++
With the growing volume of consumer-behavior data, and distributed systems being the new normal, APM solutions have evolved to actively track connections to external dependencies, like remote services. This provides an overall topology that includes all known and discovered components. The known components that have been instrumented by the APM product will provide transaction tracing between known services and track as much information as they can about discovered systems like network latency, payload size, and average processing time.
The visualization of the topology allows the APM product to expose trends. This pillar of observability shows if everything is performing as expected, or if any known components are generating errors or have below-average response times. In addition, having the topology functionality gives the ability to follow application transactions from the client, through multiple services, to the final destination. It captures and stores any errors that are generated along the way. This ability to visualize the entire infrastructure and view the full distributed trace from client transactions reduces the mean-time to recovery of an error more effectively than any other tool in the APM suite
Summary
APM solutions built on the three pillars of observability are key to gathering all the data required to manage a modern application stack and provide basic analytics capabilities. When used in conjunction with more holistic AIOps solutions, the additional machine learning capabilities will help identify deep trends and allow even more proactive enhancements to create a more robust and resilient infrastructure.
For more details on Broadcom’s APM solution, watch our recent on-demand webcast.
