Effective network troubleshooting requires collecting and correlating thousands of data points across your entire stack. The more data you ingest, however, the more data you have to search through in order to locate important signals. This can make it hard to find the information you need during time-sensitive investigations.
Datadog Network Performance Monitoring (NPM) provides enhanced search and mapping capabilities that help you quickly pinpoint critical metrics, giving you clear visibility into the health of your network at any given time. NPM’s new search experience delivers streamlined, unified filtering that enables you to comb through both client and server data simultaneously for a complete picture of communication across your network. And with clustered maps, you can organize large quantities of network data into neatly categorized visualizations to quickly sort through your endpoints.
In this post, we’ll explore how NPM helps you:
- Analyze client and server data together using a unified search bar
- Use clustered network maps to assess high-cardinality traffic
To understand the flow of traffic through your system, you need to view network metrics that show you both where that traffic originated and where it terminated. The new NPM search experience allows you to combine both client and server data in one search bar for complete network visibility and streamlined troubleshooting. You can select which type of endpoints you want to filter the data for: client, server, or a combination of the two. Then you can use features like autocomplete, recent searches, and Saved Views to create useful filter queries more easily. You can also enter the tags you want to group the data by, such as cluster, host, and region. NPM displays the results broken down by the chosen client and server groupings.
The universal search bar can help you not only quickly find the data you’re looking for but also understand what data might be missing. For example, let’s say you’re troubleshooting a recent outage involving one of your services. After filtering your client and server communications based on the relevant service tags, you discover that data isn’t being transmitted between two key hosts. Upon further investigation, you determine that this dependency was removed during a recent update and decide to contact the appropriate team to come up with a solution.
The NPM network map already helps you analyze traffic within your network by enabling you to trace the flow of requests and responses between your endpoints. By following these paths, you can easily identify dependencies and bottlenecks in your system.
To streamline the visualization experience for customers with a large number of endpoints, the network map comes with a cluster view that groups your endpoints automatically or by the attribute of your choice. You can sort your metrics by filters such as zone and environment, helping you to scope your data to the factors that are most relevant to your investigations. To dig deeper into a cluster, you can select it to access a high-level summary of performance metrics—including statistics on your volume, TCP, and connection data—for all the endpoints in that group.
You can also easily pivot to additional information about the relevant endpoints in a cluster. Let’s say that you’re troubleshooting an increase in latency within a certain region. After clustering your map based on the
datacenter tag, you can pinpoint the cluster that is experiencing the issue. From here, you can jump to a view of the related hosts in NPM. The color-coded infrastructure map then helps you identify the problematic host.
Datadog Network Performance Monitoring enables you to quickly assess your network activity, giving you real-time insights that you can use to identify root causes. With NPM’s new search experience, you can streamline your network investigations by instantly locating critical metrics and inspecting your infrastructure’s dependencies. You can then visualize those metrics in the network map view to home in on unusual activity, even for very complex networks.