Observability is a commonly used buzzword in the IT industry these days. Many companies use this term to sell their products and solutions to their customers. Though this idea could be a great selling point to many IT companies who are looking to modernize their infrastructure, it would be a challenge for these companies if they are unfamiliar with the true meaning of observability and how it will benefit their businesses moving forward.
So, what is observability? How did we get here from just the simple “monitoring”?
Observability is a distributed system’s capability to have its internal state be defined, derived, and understood based on an external source of data. This means that an observable system, such as a complex service provider network, is one that can have its status defined and understood by operators, senior management, and even customers at any given time based on the output data provided by external tools.
The more observable a network is, the quicker and more accurate one can arrive at the root cause of an existing issue.
With monitoring, the operators gather and rely on a set of data from multiple tools and provide reactive resolution. But with observability, the tools are just components that play a part in deriving the state of a system. The key here is the operators and how they can combine the output data and proactively provide the current state of a system.
For a complex, distributed system to be observable, these three pillars are key: logs, metrics, and traces. These three pillars are as important as the other and we cannot consider observability without one of all three.
3 Observability Pillars
1. LOGS
Logs are timestamped records of events that happened within the network.
Logs tell you discreetly what happened and when it happened. They provide detailed information about an event and are often presented in plaintext.
Logs are easy to generate, and most infrastructure components can provide logs that can be gathered and stored on a server or over the cloud. They are also straightforward in terms of the information it provides. The challenge for logs as a standalone set of data is that it does not provide you the high-level visibility of the network and that due to the amount of information that is contained in each log, indexing can be quite difficult.
2. METRICS
Metrics are the representations of various numerical data gathered from multiple components within the network.
With metrics, operational health, and performance data can be gathered from multiple sources and could be used as key performance indicators (KPI) of a network’s behavior. These data can be presented graphically to help in the mathematical and predictive analysis of the network’s behavior. Trends can be seen just from the metrics and alerts can be set whenever certain data exceeds or goes lower than a set threshold. Furthermore, dashboards can be customized for metrics in a way that one can view the overall stats within the network, and then dive deeper into a particular data.
Below is a sample dashboard from Accedian wherein several metrics are presented. With this, one can see the overall network health but can also navigate between other dashboards in each section to see more details of the network. For operators, this would help them in faster issue resolution.
3. TRACES
Traces can present the complete end-to-end path of a flow as it traverses throughout the network. With traces, an operator can pinpoint where an issue is happening.
In a complex, distributed system like that of a service provider, the ability to pinpoint where each traffic is coming from is critical to obtain valuable insight into a network. Having visibility of the entire paths of a flow is not just useful in debugging or troubleshooting, but it can also help prepare for capacity planning.
To read the full article, please check:
https://www.l2x.tech/observability-what-is-it-and-how-do-we-get-there/
