Infrastructure monitoring is software that watches the machines under an app: servers, disks, storage systems, network devices and cloud accounts. It raises an alarm when one of them fails, slows down or is about to run short of space or capacity. It is also called server and network monitoring, or IT infrastructure monitoring.
What does it watch? For every server, it reads how much of the processor, the memory, the disk space and the network card is in use, usually once a minute. For a storage system, it reads how full it is and how fast it answers. For a router or a switch, it reads how busy each port is, and whether any port has stopped working. It collects these numbers in three ways. It asks network devices for them over the network, using a standard method called SNMP, the Simple Network Management Protocol. On servers, a small program called an agent reports them. For machines in a public cloud, it reads the numbers that the cloud provider publishes for each account.
The main users are the IT operations teams that keep a company's machines running. Many Indian companies run a mix of both kinds of machine. Some servers sit in their own building or data centre, which is called on-premises. Others are rented from a public cloud. This mix is called a hybrid cloud, and one monitoring tool has to see both halves.
Imagine a hospital chain in Kochi with 14 hospitals. Its patient records system runs on its own servers, in its own data centre. Its appointment app runs in the cloud. One Sunday at 4 a.m., the monitoring tool notices that the disk holding the records database is 82% full, and is filling by 2% every hour. The usual rate is 0.1% a day. Nobody is awake yet, and the disk is not full. But at this rate it will be full by 1 p.m., when the outpatient departments are busiest. A full disk would stop the database, and doctors could not open any patient's records. The tool raises a warning. It also opens a ticket in the hospital's IT service management (ITSM) system, where the IT team records and tracks every problem. The engineer on duty finds that a backup job is writing its files to the wrong disk. She moves them by 8 a.m.
This early warning uses anomaly detection. The tool learns what each number normally looks like, and flags a change in its pattern before a fixed limit is crossed.
Infrastructure monitoring reads enterprise storage systems directly. When many alarms arrive from network devices at once, it passes them to network AIOps, which groups them into one cause. If the company uses an AI SRE, each alert goes to it as well.
How is it different from an observability platform? Infrastructure monitoring answers the question "is the machine healthy?" An observability platform answers "is the app's request healthy, and where did it slow down?" The two often show the same incident from two sides.