Request a quote
Emergency SectorsCareers About us Blog Get in touch
An engineer watching four screens with monitoring dashboards and graphs, with an Amsterdam canal at dusk through the window

Network monitoring: seeing the problem before the users call

Network monitoring is not the dashboard full of green dots. It is the agreement behind it: what you watch, at what value something becomes an alert, and who receives that alert. Without that, you have a screen nobody looks at and you still hear about outages from reception.

7 min read Written by the ITproposal team
Short answer

Network monitoring is about the availability and capacity of devices and links. It is not security detection and it does not measure what users experience; those are two different questions.

Start small: the internet connection, the firewall, the core switch, the wifi and the power. Ten well-tuned checks tell you more than five hundred default sensors.

An alert with no recipient is not monitoring. Agree who gets what, including out of hours, and have something outside your network check that the monitoring itself is still running.

What network monitoring is, and what it is not

Network monitoring checks continuously whether the parts of your network are reachable and coping with the load: is the internet link up, is the firewall responding, is a switch port saturating, is the DHCP pool running out of addresses. Every answer is a measurement with a timestamp, so afterwards you can also say since when something was slow.

Two other things are often mixed up with it. Security monitoring looks for attacks in logs and traffic; that is the work of a SOC and needs different sources and different people. And measuring what a user experiences (how long Teams takes to open a meeting) is digital experience monitoring. Network monitoring can tell you the line is full; it cannot tell you why one application feels slow.

The distinction matters in practice. It decides which tool you pick, who reads the alerts and what you can expect from a provider. A contract that says “monitoring” without saying which of the three promises nothing.

What to watch first

The temptation is to let the tool discover everything and switch on every sensor. Within a week you have hundreds of values and nobody knows which ones matter. Start with whatever takes a whole site down when it fails:

  • The internet connection, and the backup connection. Reachability, latency and packet loss to a fixed point outside your network. Monitor the backup line separately: it usually only gets noticed when it is needed, and then turns out to have been broken for months.
  • The firewall. Up or down, CPU and memory load, number of concurrent sessions, and the state of VPN tunnels to other sites.
  • The core switch and its uplinks. Ports flapping, error counters, and utilisation of the links between switches.
  • The wifi. Whether access points are connected to the controller, and how many clients each one is carrying.
  • Power and temperature. The UPS reports when it switches to battery; make sure that message lands somewhere. A comms cabinet running hot announces a failure days in advance.
  • Things that quietly expire. Certificates, the domain name, the firewall’s support contract. Not network traffic, but certainly network outages.

Only once that list runs quietly do you add printers, cameras and meeting rooms. In that order, not the other way round.

How it measures: ping, SNMP, flow and logs

Four techniques cover almost everything, and each answers a different question.

  • Ping and port checks: is the device alive, does the service answer. Simple and reliable, but it only says yes or no.
  • SNMP: the device reports its own state, from per-port utilisation to temperature. Use version 3, with authentication and encryption; the older versions send their community string in clear text and are still enabled by default on many devices.
  • Flow data (NetFlow, sFlow, IPFIX): who is talking to whom, and how much. That is the answer to “the line is full, but with what?”.
  • Syslog: what the device writes down itself, such as a port going down or a configuration change. Useful for reconstructing an incident afterwards.

A single-site office often gets by with the first two. Flow earns its place once you regularly have a full line without knowing why, and syslog once you want to look back at who changed what and when.

Thresholds worth an alert

The biggest risk in monitoring is not that it sees too little, but that it alerts so often that nobody looks any more. After the third false alarm in an evening, the fourth one, the real one, gets dismissed too.

Measure for a few weeks without alerts first. Then you know what normal looks like: that the line fills up with updates at nine every morning, that one particular switch runs warm at night. A threshold that is not based on that baseline is a guess.

  • Two levels: a warning that shows up in an overview, and an alert that gets someone moving. Only the second goes to a phone.
  • Dependencies: if the link to a branch drops, you want one alert, not thirty for every device behind it.
  • A delay: a port that drops once for two seconds is not an outage. Three times in fifteen minutes is.

Every quarter, go through last month’s alerts. Any alert that nobody acted on becomes a warning, or disappears.

Who gets the alert, including at 3 a.m.

An alert that lands in a mailbox nobody reads is not monitoring. For each type of alert, agree who receives it, how quickly someone picks it up, and what happens if that person does not respond.

Out of hours is the real question. Should a dead internet connection on a Saturday wake someone up? For an office that is closed at weekends, probably not; for a shop whose busiest day is Saturday, yes. That is a decision for the organisation, not for the tool, and it belongs in the agreement with your provider.

And watch the watcher. If the monitoring runs on a server in the same network, it goes down with that network and everything goes conspicuously quiet. Have something outside your network check that the monitoring is alive, and raise an alert when it falls silent.

Run your own tool, or have it done

There is no shortage of tools. PRTG from Paessler, Zabbix, Checkmk and LibreNMS are well-known names, and many network vendors offer a cloud dashboard that monitors their own devices well and everything else not at all. Choosing the tool is rarely the problem; maintaining it is.

Three questions decide whether you do it yourself:

  • Who adjusts the thresholds when the network changes, and who removes a device that has been retired?
  • Who responds when an alert comes in, and can that person actually fix it?
  • What happens to the history when the tool or the administrator changes? Last year’s measurements are the only way to see that a line is slowly becoming too small.

If the answer to the first two is “nobody in particular”, monitoring belongs with managed network services rather than with a standalone tool. In that case, ask for access to the same dashboard your provider sees, and for the right to take the measurement history with you if you leave.

Where our work stops

We set up monitoring, watch the network we manage and handle alerts the way we agree with you. Two things we can see but do not fix ourselves.

The line outside the building. The fibre, the cable or the mobile network itself belongs to the network operator. When monitoring shows the fault is there, we raise it with the provider, with the measurements attached, and follow it up. The repair is done by the operator.

Someone else’s cloud service. If Microsoft 365 or your accounting package is slow while your own network is quiet, we can demonstrate that, but we cannot fix it.

And this is about devices, not people. Tracking what individual employees do online is a different question, with data protection law and the ICO’s guidance on monitoring workers attached, and not part of network monitoring as described here.

Frequently asked

Questions we get about this

The five that come up most often.

What is network monitoring?

Network monitoring continuously checks whether the parts of a network are reachable and coping: internet connection, firewall, switches, wifi and power. When something fails or a value crosses an agreed threshold, someone receives an alert. The aim is to see a fault before a user reports it.

What is the difference between network monitoring and network management?

Monitoring is watching and alerting. Management also means acting on it: maintaining configuration, updating firmware, fixing faults and adapting the network as the organisation changes. Monitoring without anyone responding to the alerts mostly produces an archive.

Which network monitoring tools are commonly used?

Well-known tools include PRTG, Zabbix, Checkmk and LibreNMS, and many network vendors provide their own cloud dashboard. The tool matters less than its upkeep: who maintains the thresholds, who responds to alerts and where the history goes if you change provider.

Is SNMP secure enough to use?

Version 3 is, because it supports authentication and encryption. The older versions 1 and 2c send their community string in clear text and are still enabled by default on many devices. Disable them and use SNMPv3 only, restricted to the address of the monitoring server.

Do you fix faults on the internet line yourselves?

Not when the fault is in the infrastructure outside the building; that belongs to the network operator. We see the fault in the monitoring, report it to the provider with the measurements attached and follow it up until it is resolved. Anything inside the building and under our management we handle ourselves.

Related services

Further reading

What network monitoring connects to.

Brands

The brands above, one page each

What the brand makes, which lines we supply and what we do around it.

Shall we look at what your network reports today?

Tell us how you currently find out about outages and what runs at your sites. We will tell you which ten checks we would set up first and who should be receiving the alerts.

Practical IT knowledge in your inbox

New guides on management, security and the workplace, written by the people doing the work. No sales talk, and you can unsubscribe in one click.

We use your address for the newsletter only. See our privacy policy.

Request a quote Call