What is Infrastructure Monitoring?

What is infrastructure monitoring?

Infrastructure monitoring is the continuous collection and alerting on health metrics from the layer beneath your application: servers and virtual machines, containers, networks, databases and managed cloud services. Typical signals are CPU, memory, disk space and I/O, network throughput and errors, and service-specific numbers such as queue depth or replication lag. It answers one question: is the platform healthy?

How it works

A small agent or exporter runs on each host, or beside each container, and reads operating system counters. For managed services, the collector polls the cloud provider's metrics API instead. Both ship time series to a backend that stores them, draws dashboards and evaluates alert rules. Common agents include Prometheus node_exporter, the Datadog Agent, the AWS CloudWatch agent and Telegraf. Data is usually collected every few seconds to every minute, so a trend is visible long before a limit is hit.

What to measure

• CPU: utilization and run-queue length. See CPU utilization in performance testing.
• Memory: used, available and swap activity. Swapping is a stronger warning than a high "used" figure.
• Disk: free space, read and write latency, I/O wait.
• Network: bandwidth, retransmits, dropped packets and connection counts.
• Service limits: queue depth, database connections, replica lag, load balancer target health.

Example

A checkout page slows down every evening. Application metrics only say "requests are slow". The infrastructure dashboard shows the database host's disk wait climbing while CPU stays flat, and a backup job writing to the same volume. Moving the backup window fixes it. Without host-level data, the team would have blamed the code.

Common mistakes

• Alerting on averages only. An average CPU of 40% across a fleet can hide one saturated node. Alert per host and watch percentiles.
• Alerting on every spike. Short bursts are normal. Require the condition to last for a few minutes, or you train people to ignore pages.
• No baseline. A number only means something compared to what is normal for that service at that hour.
• Watching hosts, not users. Every dashboard can be green while a login endpoint returns errors.
• Forgetting managed services. Queues, caches and databases hit limits that never show up as host CPU.

Infrastructure monitoring and the rest of your stack

Infrastructure monitoring is one layer of observability. Application performance monitoring covers the code on top, network monitoring the paths between systems, and cloud monitoring the provider services you rent. It pairs well with load testing: generate controlled traffic and watch which resource saturates first, which tells you where to add capacity.

LoadFocus does not monitor hosts or containers, and it is not a replacement for an infrastructure tool. It monitors from the outside: API monitoring checks your endpoints on a schedule and alerts on failures and slow responses, and load testing produces the traffic that stresses your servers. Use an infrastructure monitor to see why, and an outside-in check to see whether users are affected. More on setting up alerts in monitoring best practices for DevOps and developers.

FAQ

What is the difference between infrastructure monitoring and application monitoring?

Infrastructure monitoring watches the platform: hosts, containers, networks and managed services. Application monitoring watches the code running on it: endpoint latency, error rates and traces. You need both, because a healthy host can still run a slow application.

Which metrics matter most?

Start with CPU, memory, disk space and I/O, and network errors for each host, then add the limits of the services you depend on, such as queue depth and database connections. Prefer metrics that show saturation, not only usage.

Do I need an agent on every server?

For virtual machines and containers you manage, usually yes, or an exporter that exposes metrics. For managed cloud services you typically read metrics through the provider's API instead.

Can LoadFocus replace my infrastructure monitoring?

No. LoadFocus checks endpoints and pages from outside and generates load for tests. It does not read CPU, memory or disk from your servers, so use it together with an infrastructure monitor.