When an application is slow, the network is often blamed first and cleared last. NPM gives network and operations teams the data to answer quickly: is a link saturated, is a router dropping packets, is the path to a cloud region adding latency, or is the problem in the application itself? Analysts often use the related term network performance monitoring and diagnostics (NPMD).
Key metrics
- Availability: whether devices and interfaces are up and reachable.
- Bandwidth utilization: how full each interface and link is.
- Latency: round-trip time between two points.
- Packet loss: the share of packets that never arrive. Even small amounts hurt TCP throughput.
- Jitter: variation in latency, which matters most for voice and video.
- Interface errors and discards: signs of bad cables, duplex mismatches or full buffers.
- Device health: CPU, memory and temperature on routers, switches and firewalls.
Where the data comes from
- SNMP polling of routers, switches and firewalls for interface counters and device health, plus SNMP traps that devices send when something changes.
- Flow records such as NetFlow, IPFIX and sFlow, which show who is talking to whom and how much. See network traffic analysis.
- Path tests (traceroute-style probes) that measure latency and loss hop by hop, including across the internet to cloud and SaaS providers.
- Packet capture for deep diagnosis of specific problems.
NPM vs. traffic analysis vs. APM
- NPM asks whether the network is delivering traffic well.
- Network traffic analysis asks what traffic is on the network and whether any of it is suspicious.
- APM asks whether the application is performing well. Seeing network and application data side by side is the quickest way to tell which one is at fault.
Good practices
- Baseline normal behavior per link and device, so real degradation stands out.
- Monitor paths, not just devices you own. Traffic to cloud and SaaS services crosses networks you don't control.
- Alert on sustained degradation rather than single spikes.
- Tie network metrics to the applications and users they affect.
Network performance management with Atatus
Atatus network monitoring uses the same Infrastructure Agent to collect SNMP device health and traps, NetFlow, IPFIX and sFlow data, and hop-by-hop path tests, and shows them next to application data. Network device monitoring covers interface traffic, errors and discards on switches, routers and firewalls, and network path monitoring measures per-hop latency and loss and detects route changes. For a comparison of tools, see network monitoring tools.