Proactive performance monitoring using New Relic AI

Managing complex technology stacks and ensuring continuous uptime are critical challenges for modern engineering teams. Reacting to issues after they occur often results in costly downtime and frustrated customers. The solution? Proactive performance monitoring. With New Relic AI, you can move from reactive to proactive monitoring by leveraging AI to foresee problems before they happen, identify trends, and resolve issues faster.

New Relic AI is the first GenAI observability assistant that combines leading large language models (LLMs) with the New Relic data platform to understand both your system and the New Relic platform itself, providing deeper insights from heaps of telemetry data using everyday language.

In this blog post, you’ll learn how to use New Relic AI to:

• Analyze performance trends
• Identify anomalies 
• Detect monitoring gaps
• Set up synthetic monitors

Getting started with NRAI

Although currently in preview, NRAI is simple to activate. To start, click Ask AI in the top right corner of any page, then click Get started. 

You will be prompted to review the Generative AI Specific Terms. If you have not previously accepted New Relic Pre-release policy, you’ll be prompted to review it here. Click Accept. Once done, you should see a chat window. 

Now you can interact with the AI assistant to help monitor and troubleshoot your system.

Analyze performance trends

Proactive monitoring starts with understanding the historical performance of your services. By analyzing trends over time, you can spot areas that might need attention before they become major issues.

To start, you can ask NRAI to summarize how a particular service has been performing over a specific period.

NRAI will retrieve relevant performance metrics, including transactions, average duration, and error rates, providing you with a quick overview of the service’s health, as shown in the example below.

In the above example, NRAI shows the performance of catalogue-service, including the transactions, average duration, and error rate. You can also ask NRAI to focus on a particular metric like error rate.

NRAI is capable of maintaining context and understands that you want to retrieve the transaction errors in "catalogue-service" during the past hours. It converts the natural language to a New Relic Query Language (NRQL) query and indicates if there were any errors in your service. 

Additionally, you can ask for a detailed explanation of these errors and also investigate their underlying causes.

Here, New Relic AI shows the list of all errors in catalogue-service along with all the associated attributes. You can go further into detail of each error. It's just a matter of asking the right question.

Another interesting point to note is that the metrics shown as part of the performance analysis vary, depending on the particular service being analyzed. For example: 
 

Now that we're asking about a browser application, the NRAI retrieves the metrics related to the browser application.

In the above example, NRAI retrieves average frontend duration, average first contentful paint, and average first paint metrics over the last 24 hours.

To ensure long-term stability, you can compare performance across different time periods. This helps highlight trends that could signal future problems.

With this, our GenAI assistant will help you see if performance is gradually declining or if spikes are becoming more frequent, which could indicate deeper issues.

In this example, the average frontend duration has improved 2%, which shows an overall improvement.

Identify anomalies

Once you've analyzed performance trends, the next step in proactive monitoring is identifying anomalies—unusual patterns or behaviors that might signal an issue. Simply ask NRAI to highlight recent anomalies. You can start by asking New Relic AI to detect any unusual activity or spikes in key metrics, such as error rates, response times, or even an overview of anomalies in your stack.

After scanning your stack, NRAI will report back any anomalies in your services, such as sudden spikes, decrease in I/O operations, error rates, or other irregular patterns.

In the above example, New Relic AI detected two anomalies: 

• Decrease in read and write operations
• Increase in write time

It also specifies the context around anomalies and shows actionable steps to resolve any related performance issues. 

Set up synthetic monitors

Synthetic monitoring simulates user interactions with your applications to detect potential issues before they affect real users. For example, NRAI can help you set up a synthetic monitor that tests the availability of your services at regular intervals. This ensures that any performance degradation is caught early. 

To start, you can ask New Relic AI for synthetic monitoring recommendations.

In the above example, NRAI scans the stack to identify the entities with anomalies and suggests setting up synthetic monitors on them. By default, NRAI creates ping checks for your services and applications. However, it can guide you to set up a more advanced synthetic monitor, such as a scripted browser or an API test. These types of monitors allow you to simulate user interactions or API calls to ensure that your application is not only reachable but also functioning correctly.

By setting up synthetic monitors, you can catch issues before they impact real users, ensuring a smooth experience for everyone interacting with your services.

Detect monitoring gaps

A crucial part of proactive monitoring is ensuring that your entire stack is covered by alerts and monitoring tools. Missing coverage in key areas could mean you're blindsided by issues. To identify unmonitored services, you can start by asking the assistant to find the gaps in your stack. Or more specifically, you can also ask for areas in your stack that are not currently covered by alerts.

Or you can also ask NRAI to find coverage gaps in your stack.

The above prompt will likely generate a list of entities that lack alert coverage, ensuring you know which parts of your system need some more attention to notify teams of issues that are currently ongoing. 

As suggested in the example, you can narrow your search by specifying the type of application you want to focus on. 

The above examples show the list of application performance monitoring (APM) services that are not covered by alert conditions. Once you’ve identified these gaps, you can prioritize setting up alerts for the most critical services first. You can also take a step further and ask New Relic AI for recommendations on which alerts should be set up based on past incidents.

This ensures that your most essential services are always monitored, helping prevent significant issues from going unnoticed.

Conclusion

New Relic AI transforms how you monitor your stack, making it easier to detect problems before they escalate. By using the AI-powered assistant for alerting, analyzing trends, investigating errors, and setting up synthetic monitors, you can shift from reactive to proactive monitoring, ensuring smoother operations and fewer incidents.